{"name":"EnvIndex company index","asOf":"2026-08-16","methodology":"https://envindex.com/methodology","semantics":{"unknown":"No reliable evidence recorded; not equivalent to No","legacyFields":"May await claim-level source migration"},"companies":[{"slug":"rise-data-labs","name":"Rise Data Labs","website":"https://risedatalabs.com","tagline":"RL environments and tasks across multiple domains, supported by a large expert network","description":"Rise Data Labs builds RL environments and tasks across multiple domains, supported by a large network of experts. Its offering includes custom environments, reward engineering, RLHF and preference data, multi-agent workflows, and adversarial testing for AI labs and enterprise teams.","domains":["computer-use","coding","finance","cybersecurity","legal","data-labeling","enterprise","multi-domain","rlhf","custom-environments"],"locations":["New York, United States"],"founders":[{"name":"Vivian M. Chen","xHandle":null}],"teamSize":"11-25","funding":null,"raising":null,"founded":null,"notable":"Runs an Automated Talent Engine of 500,000+ US professionals, 98% college-educated or higher.","entityType":"Data + environments","research":{"startupSlug":"rise-data-labs","entityType":"Data + environments","lastResearched":"2026-08-16","lastVerified":"2026-08-16","verificationStatus":"company-reported","confidence":"medium","sourceIds":["rise-data-labs-rl-environments","rise-data-labs-home","rise-data-labs-rlhf","rise-data-labs-linkedin"]},"catalogCount":3,"canonicalUrl":"https://envindex.com/startup/rise-data-labs"},{"slug":"afterquery","name":"AfterQuery","website":"https://afterquery.com","tagline":"Expert human data and RL environments across code, finance, and computer use","description":"AfterQuery recruits domain experts to produce frontier training data, computer-use trajectories, and RL-style benchmark environments spanning coding, finance, and enterprise workflows. It sells to frontier AI labs and publishes benchmarks such as AppBench.","domains":["multi-domain","coding","finance"],"locations":["San Francisco","New York","Seattle"],"founders":[{"name":"Carlos Georgescu","xHandle":"CarlosGeorgescu"},{"name":"Spencer Mateega","xHandle":"spencermateega"},{"name":"Danny Tang","xHandle":null}],"teamSize":"51-100","funding":"$30.5M total (reported)","raising":null,"founded":null,"notable":"AppBench benchmark","entityType":"Data + environments","research":{"startupSlug":"afterquery","entityType":"Data + environments","lastResearched":"2026-08-16","lastVerified":"2026-08-16","verificationStatus":"company-reported","confidence":"medium","sourceIds":["afterquery-appbench"]},"catalogCount":1,"canonicalUrl":"https://envindex.com/startup/afterquery"},{"slug":"akhara","name":"Akhara","website":"https://akhara.ai","tagline":"Enterprise and code RL environments","description":"Akhara builds enterprise and coding environments for RL training and evaluation of AI agents.","domains":["enterprise","coding"],"locations":["San Francisco"],"founders":[{"name":"Spandana Govindgari","xHandle":null},{"name":"Navi Singh","xHandle":null},{"name":"Girish Kumar","xHandle":null}],"teamSize":"1-10","funding":null,"raising":null,"founded":null,"notable":null,"entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/akhara"},{"slug":"andon-labs","name":"Andon Labs","website":"https://andonlabs.com","tagline":"Long-horizon autonomy benchmarks like Vending-Bench","description":"Andon Labs (YC, formerly Vectorview) builds benchmarks and evaluations for AI agents' long-horizon coherence and safety, including Vending-Bench, Butter-Bench, and Blueprint-Bench, and deploys agents into real-world businesses. It partnered with Anthropic on Project Vend, letting Claude run a real office vending machine.","domains":["long-horizon","alignment"],"locations":["San Francisco"],"founders":[{"name":"Lukas Petersson","xHandle":null},{"name":"Axel Backlund","xHandle":null}],"teamSize":"1-10","funding":"Seed (Y Combinator)","raising":null,"founded":null,"notable":"Anthropic's Project Vend collaboration; Vending-Bench","entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/andon-labs"},{"slug":"andromede","name":"Andromede","website":"https://andromede.ai","tagline":"Programmatic generation of long-horizon RL environments","description":"Andromede, based in Lausanne with EPFL roots, programmatically generates RL environments for long-horizon sequential reasoning tasks.","domains":["long-horizon"],"locations":["Lausanne"],"founders":[{"name":"Alexandre Sallinen","xHandle":null},{"name":"Guillaume Allegre","xHandle":null}],"teamSize":"1-10","funding":null,"raising":null,"founded":null,"notable":null,"entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/andromede"},{"slug":"anthromind","name":"Anthromind","website":"https://anthromind.com","tagline":"Medical and long-horizon environments and expert data","description":"Anthromind builds medical-domain and long-horizon environments and expert data pipelines for frontier AI training and evaluation.","domains":["medical","long-horizon","data-labeling"],"locations":["San Francisco"],"founders":[{"name":"Pratik Karki","xHandle":null},{"name":"Mannat Sandhu","xHandle":null}],"teamSize":"1-10","funding":null,"raising":null,"founded":null,"notable":null,"entityType":"Data + environments","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/anthromind"},{"slug":"applied-compute","name":"Applied Compute","website":"https://appliedcompute.com","tagline":"Ex-OpenAI trio applying RL to build specialist enterprise models","description":"Applied Compute, founded by ex-OpenAI researchers who worked on o1 and Codex, uses reinforcement learning and company-specific environments to train specialist frontier models for enterprises. It launched publicly in October 2025 with $80M raised, led by Benchmark and Sequoia.","domains":["enterprise","machine-learning","custom-environments"],"locations":["San Francisco"],"founders":[{"name":"Rhythm Garg","xHandle":null},{"name":"Linden Li","xHandle":null},{"name":"Yash Patil","xHandle":null}],"teamSize":"11-25","funding":"$80M total at $700M valuation (Oct 2025)","raising":true,"founded":2025,"notable":"Reported in talks (Jan 2026) at $1.3B valuation; founders are recent Stanford grads","entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/applied-compute"},{"slug":"arimlabs","name":"ARIMLABS","website":"https://arimlabs.ai","tagline":"Security and long-horizon environments for agentic AI","description":"ARIMLABS is a Warsaw-based team building security-focused and long-horizon RL environments, with a research background in AI agent security including published vulnerabilities in popular browser-agent frameworks.","domains":["cybersecurity","long-horizon"],"locations":["Warsaw"],"founders":[{"name":"Mykyta Mudryi","xHandle":null},{"name":"Markiian Chaklosh","xHandle":null}],"teamSize":"11-25","funding":null,"raising":null,"founded":null,"notable":"Disclosed CVEs in browser agent frameworks","entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/arimlabs"},{"slug":"artificial-analysis","name":"Artificial Analysis","website":"https://artificialanalysis.ai","tagline":"Independent benchmarking of AI models across intelligence, speed, and price","description":"Artificial Analysis independently benchmarks 575+ AI models and API providers, publishing its Intelligence Index plus agentic evals like Terminal-Bench runs, AA-Briefcase, and EnterpriseOps-Gym-AA. It is a neutral third-party evals provider rather than an environments vendor.","domains":["multi-domain","machine-learning"],"locations":["San Francisco"],"founders":[{"name":"Micah Hill-Smith","xHandle":null},{"name":"George Cameron","xHandle":null}],"teamSize":"11-25","funding":"$2.6M (2024)","raising":null,"founded":2023,"notable":"Intelligence Index; benchmarks new models within ~24 hours of release","entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/artificial-analysis"},{"slug":"benchflow","name":"BenchFlow","website":"https://benchflow.ai","tagline":"Open-source benchmark hub and eval infrastructure for agents","description":"BenchFlow builds open-source evaluation infrastructure and a hub for agent benchmarks spanning terminal, code, browser, and enterprise tasks. Its benchmarks include SkillsBench.","domains":["enterprise","browser","coding"],"locations":["San Francisco"],"founders":[{"name":"Xiangyi Li","xHandle":"xdotli"}],"teamSize":"1-10","funding":"~$1M (reported)","raising":null,"founded":null,"notable":"SkillsBench","entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/benchflow"},{"slug":"bespoke-labs","name":"Bespoke Labs","website":"https://bespokelabs.ai","tagline":"Data curation and RL environment recipes from ex-Google DeepMind researchers","description":"Bespoke Labs builds tools and curated datasets for post-training, including the Curator open-source library and RL environment work for coding and enterprise tasks. Co-founded by Mahesh Sathiamoorthy (ex-Google DeepMind) and Alex Dimakis, it co-created the OpenThoughts reasoning datasets.","domains":["coding","machine-learning"],"locations":["Mountain View","Menlo Park","Bangalore","San Francisco"],"founders":[{"name":"Mahesh Sathiamoorthy","xHandle":"madiator"},{"name":"Alex Dimakis","xHandle":"AlexGDimakis"}],"teamSize":"11-25","funding":"~$40M (reported)","raising":null,"founded":null,"notable":"OpenThoughts open reasoning datasets","entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/bespoke-labs"},{"slug":"chakra-labs","name":"Chakra Labs","website":"https://chakra.dev","tagline":"Dojo: a hub of computer-use and tool-use environments","description":"Chakra Labs builds Dojo (trydojo.ai), a computer-use environment hub for training and evaluating agents on multi-tool tasks. It sells environments to AI labs from its Brooklyn base.","domains":["computer-use","tool-use"],"locations":["Brooklyn"],"founders":[{"name":"Nirmal Krishnan","xHandle":null},{"name":"Alex Fung","xHandle":null}],"teamSize":"11-25","funding":"~$10.1M (reported)","raising":null,"founded":null,"notable":"Dojo environment platform","entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/chakra-labs"},{"slug":"collinear","name":"Collinear","website":"https://collinear.ai","tagline":"Enterprise simulation, judges, and long-horizon trajectory generation","description":"Collinear AI, founded by former Hugging Face research lead Nazneen Rajani, builds LLM judges, evaluation curators, and enterprise workflow simulation environments for post-training and RL. It sells trajectory generation and reliability tooling to labs and enterprises.","domains":["enterprise","long-horizon","machine-learning","simulation"],"locations":["Mountain View","Sunnyvale"],"founders":[{"name":"Nazneen Rajani","xHandle":"nazneenrajani"},{"name":"Soumyadeep Bakshi","xHandle":null}],"teamSize":"11-25","funding":null,"raising":null,"founded":2023,"notable":"Curator Evals","entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/collinear"},{"slug":"datacurve","name":"Datacurve","website":"https://datacurve.ai","tagline":"Frontier coding data and repository RL environments via the Shipd bounty platform","description":"Datacurve (YC W24) supplies high-quality, complex coding data, RLHF traces, and repository-based RL environments to foundation model labs. Its Shipd platform gamifies data collection with bounties completed by 1,400+ vetted software engineers.","domains":["coding","rlhf"],"locations":["San Francisco"],"founders":[{"name":"Serena Ge","xHandle":"serenaa_ge"},{"name":"Charley Lee","xHandle":null}],"teamSize":"26-50","funding":"Series A, $15M led by Chemistry (Oct 2025); $17.7M total","raising":null,"founded":2024,"notable":"DeepSWE benchmark work","entityType":"Data + environments","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/datacurve"},{"slug":"deeptune","name":"Deeptune","website":"https://deeptune.com","tagline":"Code and computer-use environments; acquired by Mercor","description":"Deeptune built coding and computer-use RL environments and enterprise workflow simulation for AI labs. It raised a $43M Series A and was acquired by Mercor in July 2026.","domains":["coding","computer-use"],"locations":["New York"],"founders":[{"name":"Tim Lupo","xHandle":null}],"teamSize":"26-50","funding":"Series A, $43M; acquired by Mercor (2026)","raising":false,"founded":null,"notable":"Acquired by Mercor in July 2026","entityType":"Acquired / inactive","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/deeptune"},{"slug":"diffuse-labs","name":"Diffuse Labs","website":"https://diffuselabs.ai","tagline":"ML and long-horizon RL environments","description":"Diffuse Labs builds machine-learning and long-horizon environments for RL training of frontier models.","domains":["machine-learning","long-horizon"],"locations":["Palo Alto","San Francisco"],"founders":[{"name":"Daljeet Virdi","xHandle":null},{"name":"Tejpal Singh","xHandle":null}],"teamSize":"1-10","funding":null,"raising":null,"founded":null,"notable":null,"entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/diffuse-labs"},{"slug":"dissei","name":"Dissei","website":"https://dissei.ai","tagline":"Finance-domain RL environments","description":"Dissei is a London-based startup building finance-domain environments for training and evaluating AI models on financial workflows.","domains":["finance"],"locations":["London"],"founders":[{"name":"Tudor Popescu","xHandle":null},{"name":"Jai Trivedi","xHandle":null},{"name":"Eddy B","xHandle":null}],"teamSize":"1-10","funding":null,"raising":null,"founded":null,"notable":null,"entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/dissei"},{"slug":"edotenv","name":"EdotEnv","website":"https://edotenv.com","tagline":"Long-horizon planning environments for frontier models","description":"EdotEnv builds long-horizon and machine-learning RL environments, with a focus on long-horizon planning tasks for frontier labs.","domains":["long-horizon","machine-learning"],"locations":["San Francisco"],"founders":[{"name":"Rui Wang","xHandle":null},{"name":"Michael Zhang","xHandle":null}],"teamSize":"1-10","funding":null,"raising":null,"founded":null,"notable":null,"entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/edotenv"},{"slug":"epoch-ai","name":"Epoch AI","website":"https://epoch.ai","tagline":"Nonprofit research institute behind FrontierMath and AI capability benchmarks","description":"Epoch AI is a nonprofit research institute that studies AI trends and builds rigorous benchmarks, including FrontierMath and the Epoch Capabilities Index. Not a startup selling environments, but a key independent evals organization; Mechanize's founders previously led Epoch.","domains":["math","machine-learning"],"locations":["Remote"],"founders":[{"name":"Jaime Sevilla","xHandle":null},{"name":"Tamay Besiroglu","xHandle":"tamaybes"}],"teamSize":"11-25","funding":"Philanthropic grants (nonprofit)","raising":null,"founded":2022,"notable":"FrontierMath benchmark commissioned by OpenAI","entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/epoch-ai"},{"slug":"exabite","name":"Exabite","website":"https://exabite.ai","tagline":"Code RL environments with realistic software execution","description":"Exabite builds code-centric RL environments with a focus on realistic software execution, selling to frontier AI labs.","domains":["coding"],"locations":["Remote"],"founders":[{"name":"Alana Xiang","xHandle":null}],"teamSize":"1-10","funding":null,"raising":null,"founded":null,"notable":null,"entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/exabite"},{"slug":"general-reasoning","name":"General Reasoning","website":"https://gr.inc","tagline":"Open reasoning data and reward models from the ex-Meta AI reasoning lead","description":"General Reasoning, co-founded by Ross Taylor (former Meta AI reasoning/Galactica lead), builds open datasets, reward models, and long-horizon RL infrastructure, with finance-domain environments. It runs the OpenReward project.","domains":["finance","long-horizon","machine-learning"],"locations":["London","San Francisco"],"founders":[{"name":"Ross Taylor","xHandle":"rosstaylor90"},{"name":"Chengxi Taylor","xHandle":null},{"name":"Kip Parker","xHandle":null},{"name":"Thomas Grady","xHandle":null}],"teamSize":"1-10","funding":"~$10.9M (reported)","raising":null,"founded":null,"notable":"OpenReward; founder co-created Galactica and Llama reasoning work at Meta","entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/general-reasoning"},{"slug":"good-start-labs","name":"Good Start Labs","website":"https://goodstartlabs.com","tagline":"Game-based RL environments and benchmarks","description":"Good Start Labs builds game-like and long-horizon environments for training and benchmarking AI models. Co-founder Alex Duffy created the AI Diplomacy benchmark, which pits frontier models against each other in the strategy game Diplomacy.","domains":["games","long-horizon"],"locations":["Brooklyn","New York","Toronto"],"founders":[{"name":"Alex Duffy","xHandle":null},{"name":"Tyler Marques","xHandle":null}],"teamSize":"1-10","funding":"~$3.6M (reported)","raising":null,"founded":null,"notable":"AI Diplomacy benchmark","entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/good-start-labs"},{"slug":"gray-swan-ai","name":"Gray Swan AI","website":"https://grayswan.ai","tagline":"Adversarial red-teaming arenas and safety evals for frontier models","description":"Gray Swan AI, founded by CMU professors Zico Kolter and Matt Fredrikson with Andy Zou, runs adversarial red-teaming arenas, jailbreak competitions, and security evaluations used by frontier labs including OpenAI and Anthropic. Its environments stress-test agent robustness and safety.","domains":["cybersecurity","alignment"],"locations":["Pittsburgh"],"founders":[{"name":"Zico Kolter","xHandle":"zicokolter"},{"name":"Matt Fredrikson","xHandle":null},{"name":"Andy Zou","xHandle":null}],"teamSize":"11-25","funding":"~$40M (reported)","raising":null,"founded":2023,"notable":"Co-founder Zico Kolter chairs OpenAI's safety committee; ran red-teaming arenas for GPT and Claude launches","entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/gray-swan-ai"},{"slug":"halluminate","name":"Halluminate","website":"https://halluminate.ai","tagline":"Sandboxed RL environments for finance and enterprise workflows","description":"Halluminate builds sandboxed RL environments and evals with a focus on financial services and enterprise knowledge work. It publishes WebBench, a benchmark for browser agents.","domains":["finance","enterprise","browser"],"locations":["San Francisco"],"founders":[{"name":"Jerry Wu","xHandle":null},{"name":"Wyatt Marshall","xHandle":null}],"teamSize":"26-50","funding":null,"raising":null,"founded":null,"notable":"WebBench browser-agent benchmark","entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/halluminate"},{"slug":"handshake","name":"Handshake","website":"https://joinhandshake.com/research/ai","tagline":"Career network turned human-data and RL environments provider via Handshake AI","description":"Handshake operates the largest early-career network in the US and launched Handshake AI, a human data labs business that recruits PhD-level experts to build evals and RL environments for frontier AI labs. It leverages its network of millions of students and experts for domain-specific data.","domains":["multi-domain","data-labeling","rlhf"],"locations":["San Francisco","New York","Bangalore","Berlin"],"founders":[{"name":"Garrett Lord","xHandle":null}],"teamSize":"250+","funding":"Series F, $200M (2022, ~$3.5B valuation)","raising":null,"founded":2014,"notable":"Known for its 'Gandalf the Grader' benchmark work","entityType":"Data + environments","research":{"startupSlug":"handshake","entityType":"Data + environments","lastResearched":"2026-08-16","lastVerified":"2026-08-16","verificationStatus":"confirmed","confidence":"high","sourceIds":["handshake-gandalf","handshake-gandalf-github","handshake-banker-dataset"]},"catalogCount":2,"canonicalUrl":"https://envindex.com/startup/handshake"},{"slug":"hillclimb","name":"Hillclimb","website":"https://hillclimb.com","tagline":"Math environments emphasizing verifiable correctness","description":"Hillclimb builds mathematics-domain environments emphasizing correctness and verifiable rewards for frontier model training.","domains":["math"],"locations":["San Francisco"],"founders":[{"name":"Jun Park","xHandle":null},{"name":"Ibrakhim Ustelbay","xHandle":null}],"teamSize":"1-10","funding":null,"raising":null,"founded":null,"notable":null,"entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/hillclimb"},{"slug":"hud","name":"HUD","website":"https://hud.ai","tagline":"Evals and RL environments platform for computer-use agents","description":"HUD (YC W25) provides a platform for building agent evals and RL environments: teams wrap real software as agent-callable tools in isolated containers, define tasks and rewards, and run evals/RL at scale. Over 50 businesses and frontier labs use it, with 2,500+ environments built and 1.3M+ task runs.","domains":["computer-use","coding","long-horizon","agents-infrastructure","custom-environments"],"locations":["San Francisco","Singapore"],"founders":[{"name":"Lorenss Martinsons","xHandle":null},{"name":"Jay Ram","xHandle":null}],"teamSize":"11-25","funding":"$15M raised (YC W25, Exceptional Capital)","raising":null,"founded":null,"notable":"DoorDash and UiPath among users; hosts hosted versions of benchmarks like OSWorld","entityType":"Environment platform","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/hud"},{"slug":"huzzle-labs","name":"Huzzle Labs","website":"https://labs.huzzle.com","tagline":"Long-horizon code, tool-use, and enterprise workflow environments","description":"Huzzle Labs, spun out of the Huzzle career platform, builds long-horizon coding, tool-use, and enterprise workflow RL environments for AI labs.","domains":["long-horizon","coding","enterprise"],"locations":["London","Berlin","San Francisco"],"founders":[{"name":"Ingmar Klein","xHandle":null},{"name":"Amit Choudhary","xHandle":null}],"teamSize":"26-50","funding":"~$6M (reported)","raising":null,"founded":null,"notable":null,"entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/huzzle-labs"},{"slug":"idler","name":"Idler","website":"https://idler.ai","tagline":"Code environments with realistic execution constraints","description":"Idler builds code-centric RL environments emphasizing realistic software execution constraints for training coding agents.","domains":["coding"],"locations":["San Francisco"],"founders":[{"name":"Ivan Chub","xHandle":null},{"name":"Nalu Concepcion","xHandle":null},{"name":"Tony Goss","xHandle":null}],"teamSize":"11-25","funding":null,"raising":null,"founded":null,"notable":null,"entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/idler"},{"slug":"incalmo","name":"Incalmo","website":"https://incalmo.ai","tagline":"Offensive-security environments for AI agents","description":"Incalmo, spun out of Carnegie Mellon research, builds cybersecurity environments and tooling for evaluating and training AI agents on network attack/defense scenarios. Its research on LLM-orchestrated network intrusions preceded the company.","domains":["cybersecurity"],"locations":["San Mateo"],"founders":[{"name":"Brian Singer","xHandle":null},{"name":"Mark Dong","xHandle":null}],"teamSize":"11-25","funding":null,"raising":null,"founded":null,"notable":"CMU spinout; research on autonomous network attack agents","entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/incalmo"},{"slug":"invariant-labs","name":"Invariant Labs","website":"https://invariantlabs.ai","tagline":"Security testing and analysis for AI agents; acquired by Snyk","description":"Invariant Labs, an ETH Zurich spin-off, built tools for testing, tracing, and securing AI agents, discovering attack classes like MCP tool poisoning and rug pulls. It was acquired by Snyk in June 2025 to anchor Snyk's agentic AI security research.","domains":["cybersecurity","alignment"],"locations":["Zurich"],"founders":[{"name":"Marc Fischer","xHandle":null},{"name":"Luca Beurer-Kellner","xHandle":null}],"teamSize":"1-10","funding":"Acquired by Snyk (June 2025)","raising":false,"founded":2024,"notable":"Coined 'tool poisoning' and 'MCP rug pull' attacks","entityType":"Acquired / inactive","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/invariant-labs"},{"slug":"latch","name":"Latch","website":"https://latch.bio","tagline":"Biology data infrastructure turned life-science environments and benchmarks","description":"LatchBio provides cloud infrastructure for biocomputing and has extended into biology-domain benchmarks and environments for AI models, publishing benchmarks.bio. It applies its wet-lab and bioinformatics platform to grade and train scientific AI agents.","domains":["science"],"locations":["San Francisco"],"founders":[{"name":"Alfredo Andere","xHandle":null},{"name":"Kenny Workman","xHandle":null},{"name":"Kyle Giffin","xHandle":null}],"teamSize":"11-25","funding":"Series A, $15M (2022)","raising":null,"founded":2021,"notable":"benchmarks.bio","entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/latch"},{"slug":"lmarena","name":"LMArena","website":"https://lmarena.ai","tagline":"Crowdsourced model leaderboards from the Chatbot Arena team","description":"LMArena, spun out of UC Berkeley's Chatbot Arena project, runs crowdsourced head-to-head AI model evaluations and leaderboards used across the industry. Evals-focused rather than RL environments, but a core evaluation vendor to labs.","domains":["multi-domain","machine-learning"],"locations":["San Francisco","Berkeley"],"founders":[{"name":"Anastasios Angelopoulos","xHandle":"ml_angelopoulos"},{"name":"Wei-Lin Chiang","xHandle":"infwinston"},{"name":"Ion Stoica","xHandle":null}],"teamSize":"26-50","funding":"$100M seed (a16z, UC Investments, 2025); $150M at $1.7B valuation (Jan 2026)","raising":null,"founded":2024,"notable":"Chatbot Arena leaderboard is a de facto industry standard","entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/lmarena"},{"slug":"matrices","name":"Matrices","website":"https://matrices.ai","tagline":"Browser-native training environments for web agents","description":"Matrices builds browser and computer-use training environments for web navigation RL and evaluation.","domains":["browser","computer-use"],"locations":["San Francisco"],"founders":[{"name":"Leonardo Axel Setyanto","xHandle":null},{"name":"John Qian","xHandle":null}],"teamSize":"1-10","funding":"~$5M (reported)","raising":null,"founded":null,"notable":null,"entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/matrices"},{"slug":"mechanize","name":"Mechanize","website":"https://mechanize.work","tagline":"RL environments to automate software engineering, founded by ex-Epoch AI researchers","description":"Mechanize builds RL environments and 'boot camps' aimed at automating white-collar work, starting with software engineering agents. Founded by former Epoch AI leaders, it works with Anthropic on RL environments and is backed by investors including Nat Friedman, Daniel Gross, and Patrick Collison.","domains":["coding"],"locations":["San Francisco"],"founders":[{"name":"Tamay Besiroglu","xHandle":"tamaybes"},{"name":"Ege Erdil","xHandle":"EgeErdil2"},{"name":"Matthew Barnett","xHandle":"MatthewJBar"}],"teamSize":"51-100","funding":"~$9.1M (reported)","raising":null,"founded":2025,"notable":"Offered $500K salaries to engineers building environments; publishes GBA Eval","entityType":"Pure-play commercial","research":{"startupSlug":"mechanize","entityType":"Pure-play commercial","lastResearched":"2026-08-16","lastVerified":"2026-08-16","verificationStatus":"company-reported","confidence":"medium","sourceIds":["mechanize-gba"]},"catalogCount":1,"canonicalUrl":"https://envindex.com/startup/mechanize"},{"slug":"mercor","name":"Mercor","website":"https://mercor.com","tagline":"Expert marketplace powering evals and RL environments for frontier labs","description":"Mercor connects AI labs with vetted domain experts for model training, evaluations, and RL environments across coding, healthcare, law, and other domains. Customers include OpenAI, Meta, and Anthropic, and it reached a ~$450M run rate in 2025. It has acquired RL-environment startups Sepal AI and Deeptune.","domains":["multi-domain","data-labeling","rlhf"],"locations":["San Francisco"],"founders":[{"name":"Brendan Foody","xHandle":"BrendanFoody"},{"name":"Adarsh Hiremath","xHandle":null},{"name":"Surya Midha","xHandle":null}],"teamSize":"250+","funding":"Series C, $350M at $10B valuation (Oct 2025)","raising":true,"founded":2023,"notable":"APEX benchmark; reported in talks (July 2026) to raise $500M at $20B valuation; founders became youngest self-made billionaires","entityType":"Incumbent","research":{"startupSlug":"mercor","entityType":"Incumbent","lastResearched":"2026-08-16","lastVerified":"2026-08-16","verificationStatus":"confirmed","confidence":"high","sourceIds":["mercor-apex","mercor-apex-data"]},"catalogCount":1,"canonicalUrl":"https://envindex.com/startup/mercor"},{"slug":"metaphi","name":"Metaphi","website":"https://metaphi.ai","tagline":"Code and enterprise RL environments","description":"Metaphi builds coding and enterprise environments for training and evaluating AI agents, including its CREW benchmark.","domains":["coding","enterprise"],"locations":["San Francisco","New York"],"founders":[{"name":"Abhishek Chandwani","xHandle":null},{"name":"Ishan Gupta","xHandle":null}],"teamSize":"1-10","funding":null,"raising":null,"founded":null,"notable":"CREW benchmark","entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/metaphi"},{"slug":"micro1","name":"Micro1","website":"https://micro1.ai","tagline":"Vetted domain experts for AI training data and evals","description":"Micro1 provides vetted human experts (PhDs, senior engineers) to AI labs for model training data, evaluations, and RL-related human data work, positioning itself as a Scale AI challenger. It crossed $100M annualized revenue in 2025 with clients including Microsoft.","domains":["data-labeling","multi-domain","rlhf"],"locations":["Los Angeles"],"founders":[{"name":"Ali Ansari","xHandle":null}],"teamSize":"101-250","funding":"Series A, $35M at $500M valuation (Sept 2025)","raising":null,"founded":2022,"notable":"Fielded offers at $2.5B valuation by Dec 2025 per Forbes","entityType":"Data + environments","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/micro1"},{"slug":"normal","name":"Normal","website":"https://normal.ai","tagline":"Hardware engineering environments for AI models","description":"Normal builds hardware-engineering environments and evals, training AI models on physical/electrical engineering tasks. Co-founder HudZah is known for building a nuclear fusor with Claude's help.","domains":["hardware-engineering"],"locations":["San Francisco"],"founders":[{"name":"HudZah","xHandle":null},{"name":"Anson Yu","xHandle":null}],"teamSize":"1-10","funding":null,"raising":null,"founded":null,"notable":"Co-founder built a nuclear fusor in his kitchen using Claude","entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/normal"},{"slug":"nous-research","name":"Nous Research","website":"https://nousresearch.com","tagline":"Open AI lab behind the Atropos RL environments framework","description":"Nous Research is an open-source AI lab known for the Hermes model family and Atropos, its open rollout framework for LLM reinforcement learning environments spanning code execution, tool calling, and multimodal tasks. It also builds Psyche for distributed training.","domains":["machine-learning","agents-infrastructure"],"locations":["New York"],"founders":[{"name":"Karan Malhotra","xHandle":"karan4d"},{"name":"Jeffrey Quesnelle","xHandle":"theemozilla"},{"name":"Teknium","xHandle":"Teknium1"}],"teamSize":"11-25","funding":"Series A, $50M led by Paradigm (~$1B valuation, 2025)","raising":null,"founded":2023,"notable":"Atropos open-source RL environments framework","entityType":"Environment platform","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/nous-research"},{"slug":"osmosis","name":"Osmosis","website":"https://osmosis.ai","tagline":"Forward-deployed reinforcement learning for AI agents","description":"Osmosis (YC W25) is a post-training platform that fine-tunes models with reinforcement learning so companies' agents beat foundation models on specific tasks in performance, cost, and latency. Founded by Kasey Zhang and ex-TikTok tech lead Andy Lyu.","domains":["machine-learning"],"locations":["San Francisco"],"founders":[{"name":"Kasey Zhang","xHandle":null},{"name":"Andy Lyu","xHandle":null}],"teamSize":"1-10","funding":"Seed, $7M (CRV, Audacious Ventures, YC)","raising":null,"founded":2024,"notable":"Angel investors include Paul Graham and Guillermo Rauch","entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/osmosis"},{"slug":"pareto","name":"Pareto","website":"https://pareto.ai","tagline":"Expert data workforce for RLHF, evals, and tool-use environments","description":"Pareto AI runs a managed expert workforce producing labeled data, RLHF, and evaluation/environment work for AI labs across many domains. It is behind the Leap and Attune benchmark efforts.","domains":["multi-domain","tool-use","data-labeling","rlhf"],"locations":["San Francisco"],"founders":[{"name":"Phoebe Yao","xHandle":null}],"teamSize":"26-50","funding":null,"raising":null,"founded":null,"notable":"Leap and Attune benchmarks","entityType":"Data + environments","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/pareto"},{"slug":"plato","name":"Plato","website":"https://plato.so","tagline":"High-fidelity replicas of websites and software for agent training","description":"Plato builds replica environments of real websites and enterprise software so browser and computer-use agents can be trained and evaluated safely with verifiable rewards.","domains":["browser","enterprise","simulation"],"locations":["San Francisco"],"founders":[{"name":"Rob Farlow","xHandle":null},{"name":"Pranav Putta","xHandle":null}],"teamSize":"1-10","funding":null,"raising":null,"founded":null,"notable":null,"entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/plato"},{"slug":"pre-dev","name":"pre.dev","website":"https://pre.dev/rl-environments","tagline":"Software-planning platform offering coding and long-horizon RL environments","description":"pre.dev builds product-spec and architecture generation tooling and offers RL environments for coding and long-horizon software tasks, reporting results on Terminal-Bench 2.","domains":["coding","long-horizon"],"locations":["Delaware"],"founders":[{"name":"Arjun Raj Jain","xHandle":null},{"name":"Adam Elkassas","xHandle":null}],"teamSize":"1-10","funding":null,"raising":null,"founded":null,"notable":"Terminal-Bench 2 results","entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/pre-dev"},{"slug":"preference-model","name":"Preference Model","website":"https://preferencemodel.com","tagline":"Stealth startup working on preference and reward modeling","description":"Preference Model is a stealth-stage company working on preference/reward-modeling signal quality and code environments for frontier training.","domains":["machine-learning","coding"],"locations":["San Francisco","Toronto","Seattle"],"founders":[],"teamSize":"11-25","funding":null,"raising":null,"founded":null,"notable":null,"entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/preference-model"},{"slug":"prime-intellect","name":"Prime Intellect","website":"https://primeintellect.ai","tagline":"Open superintelligence stack: compute, RL environments hub, and sandboxes","description":"Prime Intellect provides compute, large-scale RL training, environments, sandboxes, and evals so anyone can train and improve their own AI agents. Its open Environments Hub, described as a 'Hugging Face for RL environments', hosts thousands of community environments, and it trained the INTELLECT model series with large-scale RL.","domains":["machine-learning","agents-infrastructure","multi-domain"],"locations":["San Francisco"],"founders":[{"name":"Vincent Weisser","xHandle":"vincentweisser"},{"name":"Johannes Hagemann","xHandle":"johannes_hage"}],"teamSize":"26-50","funding":"Series A, $130M at $1B valuation (2026); $150M+ total","raising":null,"founded":2024,"notable":"Backed by Founders Fund, Menlo, Radical Ventures, NVIDIA, and Andrej Karpathy; $100M+ annualized revenue","entityType":"Environment platform","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/prime-intellect"},{"slug":"proximal","name":"Proximal","website":"https://proximal.ai","tagline":"Long-horizon coding RL environments built from real codebases","description":"Proximal builds long-horizon software engineering environments from real codebases for RL training of coding agents at frontier labs. It publishes the FrontierSWE benchmark.","domains":["coding","long-horizon"],"locations":["San Francisco","Bangalore"],"founders":[{"name":"Justus Mattern","xHandle":null},{"name":"Navid Pour","xHandle":null},{"name":"Calvin Chen","xHandle":null}],"teamSize":"26-50","funding":null,"raising":null,"founded":null,"notable":"FrontierSWE benchmark","entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/proximal"},{"slug":"quesma","name":"Quesma","website":"https://quesma.com","tagline":"Security-domain RL environments and binary analysis evals","description":"Quesma, a Warsaw-based team that originally built a database gateway, now builds security-focused RL environments and evaluations for AI agents, including binary auditing tasks. It publishes the BinaryAudit benchmark.","domains":["cybersecurity","coding"],"locations":["Warsaw"],"founders":[{"name":"Jacek Migdal","xHandle":null}],"teamSize":"11-25","funding":null,"raising":null,"founded":null,"notable":"BinaryAudit benchmark","entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/quesma"},{"slug":"reasoncore","name":"ReasonCore","website":"https://reasoncore.ai","tagline":"Science and code reasoning environments and benchmarks","description":"ReasonCore builds environments and benchmarks targeting scientific and coding reasoning for training and evaluating frontier models.","domains":["science","coding"],"locations":["San Francisco"],"founders":[{"name":"Geoff Wolfe","xHandle":null},{"name":"Anshuman Lall","xHandle":null},{"name":"Satish Vutukuru","xHandle":null}],"teamSize":"11-25","funding":null,"raising":null,"founded":null,"notable":null,"entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/reasoncore"},{"slug":"refresh","name":"Refresh","website":"https://refresh.dev","tagline":"Simulation engines with verifiable rewards for coding and computer use","description":"Refresh builds coding and computer-use simulation engines with verifiable reward signals for RL training of software agents.","domains":["coding","computer-use","simulation"],"locations":["San Francisco"],"founders":[{"name":"Christopher Settles","xHandle":null},{"name":"Erik Quintanilla","xHandle":null}],"teamSize":"1-10","funding":null,"raising":null,"founded":null,"notable":null,"entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/refresh"},{"slug":"scale","name":"Scale","website":"https://scale.com","tagline":"Data-labeling incumbent extending into agent evals and RL environments","description":"Scale AI is the original data-labeling powerhouse for AI labs and enterprises, now building RL environments and agent evaluation products under its agents and RL environments group. After Meta's 2025 investment and the departure of CEO Alexandr Wang, it lost some lab customers but continues to push into environments.","domains":["multi-domain","coding","data-labeling","rlhf"],"locations":["San Francisco","New York","Washington DC","London"],"founders":[{"name":"Alexandr Wang","xHandle":"alexandr_wang"},{"name":"Lucy Guo","xHandle":null}],"teamSize":"250+","funding":"$1.6B+ raised; Meta invested $14.3B at ~$29B valuation (June 2025)","raising":null,"founded":2016,"notable":"Publishes SWE-bench Pro benchmark","entityType":"Incumbent","research":{"startupSlug":"scale","entityType":"Incumbent","lastResearched":"2026-08-16","lastVerified":"2026-08-16","verificationStatus":"confirmed","confidence":"high","sourceIds":["scale-swe-pro","scale-swe-public"]},"catalogCount":2,"canonicalUrl":"https://envindex.com/startup/scale"},{"slug":"sepal-ai","name":"Sepal AI","website":"https://sepalai.com","tagline":"Science-domain environments; acquired by Mercor","description":"Sepal AI built science-oriented environments and expert data for research workflows before being acquired by Mercor in February 2026.","domains":["science"],"locations":["San Francisco"],"founders":[{"name":"Robi Lin","xHandle":null},{"name":"Kat Hu","xHandle":null}],"teamSize":"1-10","funding":"Acquired by Mercor (Feb 2026)","raising":false,"founded":null,"notable":"Acquired by Mercor","entityType":"Acquired / inactive","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/sepal-ai"},{"slug":"snorkel","name":"Snorkel","website":"https://snorkel.ai/research","tagline":"Programmatic data platform expanding into expert evals and RL environments","description":"Snorkel AI, born from the Stanford AI Lab's Snorkel project, provides programmatic data labeling and an expert data service, and now builds agent evaluations and RL environments for frontier model developers. Its research arm publishes benchmarks including a senior software engineering eval.","domains":["multi-domain","coding","machine-learning","data-labeling","rlhf"],"locations":["San Francisco","Redwood City","New York"],"founders":[{"name":"Alex Ratner","xHandle":null},{"name":"Christopher Ré","xHandle":null},{"name":"Paroma Varma","xHandle":null},{"name":"Henry Ehrenberg","xHandle":null}],"teamSize":"101-250","funding":"Series D, $100M at $1.3B valuation (2025)","raising":null,"founded":2019,"notable":"Senior SWE-Bench benchmark work","entityType":"Data + environments","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/snorkel"},{"slug":"surge","name":"Surge","website":"https://surgehq.ai","tagline":"Bootstrapped human-data leader with a dedicated RL environments org","description":"Surge AI is the largest human data / RLHF vendor by revenue ($1.2B+ in 2024) serving OpenAI, Google, Anthropic, and Meta, and has created an internal organization dedicated to RL environments. It builds expert-graded evals and environment products such as EnterpriseBench.","domains":["multi-domain","data-labeling","rlhf"],"locations":["San Francisco","New York","Seattle"],"founders":[{"name":"Edwin Chen","xHandle":null}],"teamSize":"101-250","funding":"Bootstrapped; reported in talks to raise ~$1B at $25B+ valuation (2025)","raising":true,"founded":2020,"notable":"~$9.9M revenue per employee; CoreCraft benchmark","entityType":"Incumbent","research":{"startupSlug":"surge","entityType":"Incumbent","lastResearched":"2026-08-16","lastVerified":"2026-08-16","verificationStatus":"company-reported","confidence":"high","sourceIds":["surge-products","surge-off-the-shelf","surge-corecraft","surge-benchmarks"]},"catalogCount":11,"canonicalUrl":"https://envindex.com/startup/surge"},{"slug":"synthlabs","name":"SynthLabs","website":"https://synthlabs.ai","tagline":"Post-training research: synthetic data and scalable RL alignment","description":"SynthLabs is a post-training research company combining RLHF with synthetic AI feedback (RLAIF) to build scalable alignment and reasoning pipelines, publishing open research and datasets used in RL post-training.","domains":["machine-learning","alignment","rlhf"],"locations":["San Francisco"],"founders":[{"name":"Louis Castricato","xHandle":null},{"name":"Nathan Lile","xHandle":null}],"teamSize":"1-10","funding":"Seed (M12 and First Spark Ventures, 2024)","raising":null,"founded":2023,"notable":"Team includes founders of the trlX RLHF library","entityType":"Data + environments","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/synthlabs"},{"slug":"tacit-labs","name":"Tacit Labs","website":"https://tacitlabs.co","tagline":"Life-science and long-horizon environments for AI models","description":"Tacit Labs builds science-domain environments and benchmarks capturing tacit expert knowledge, including LifeSciBench for life-science tasks.","domains":["science","long-horizon"],"locations":["San Francisco"],"founders":[{"name":"Nicole Fitzgerald","xHandle":null},{"name":"Anne Marie Droste","xHandle":null}],"teamSize":"1-10","funding":null,"raising":null,"founded":null,"notable":"LifeSciBench","entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/tacit-labs"},{"slug":"taste-labs","name":"Taste Labs","website":"https://tastelabs.com","tagline":"Design-domain environments and evals for AI models","description":"Taste Labs builds design-focused environments and human-preference evaluations to train models on aesthetic and design tasks.","domains":["design"],"locations":["New York"],"founders":[{"name":"Thais Castello Branco","xHandle":null}],"teamSize":"26-50","funding":null,"raising":null,"founded":null,"notable":null,"entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/taste-labs"},{"slug":"trajectory-labs","name":"Trajectory Labs","website":"https://trajectorylabs.com","tagline":"Alignment-focused environments for safe agent trajectories","description":"Trajectory Labs works on alignment-focused environment design aimed at producing safe and robust agent trajectories.","domains":["alignment"],"locations":["Berkeley","Toronto"],"founders":[{"name":"Peter McIntyre","xHandle":null},{"name":"Ben West","xHandle":null}],"teamSize":"1-10","funding":null,"raising":null,"founded":null,"notable":null,"entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/trajectory-labs"},{"slug":"turing","name":"Turing","website":"https://turing.com/advance/rl-environments","tagline":"AGI infrastructure: coding data and RL environments at scale","description":"Turing began as a talent marketplace for remote software engineers and pivoted into 'AGI infrastructure', supplying coding data, model evaluations, and RL environments to frontier AI labs. Its RL environments offering targets code, reasoning, and multimodal tasks.","domains":["coding","multi-domain","data-labeling","rlhf"],"locations":["San Francisco","Palo Alto","Gurugram"],"founders":[{"name":"Jonathan Siddharth","xHandle":null},{"name":"Vijay Krishnan","xHandle":null}],"teamSize":"250+","funding":"Series E, $111M at $2.2B valuation (2025)","raising":null,"founded":2018,"notable":"Known for its VLMBench work","entityType":"Data + environments","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/turing"},{"slug":"ulam","name":"Ulam","website":"https://ulam.ai","tagline":"Math RL environments and RLVR trajectories","description":"Ulam, founded by math PhD and AI entrepreneur Przemek Chojecki, produces mathematics environments and reinforcement learning with verifiable rewards (RLVR) trajectory data for frontier labs.","domains":["math"],"locations":["Warsaw","London"],"founders":[{"name":"Przemek Chojecki","xHandle":null}],"teamSize":"1-10","funding":null,"raising":null,"founded":null,"notable":"RLVR trajectory datasets","entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/ulam"},{"slug":"vals-ai","name":"Vals AI","website":"https://vals.ai","tagline":"Independent domain-specific benchmarks for legal, finance, and tax AI","description":"Vals AI is a third-party evaluation company that builds domain-specific benchmarks with practicing professionals, measuring how LLMs perform on real legal, tax, finance, and healthcare tasks. Its Vals Legal AI Report, built with a consortium of law firms, is an industry reference.","domains":["legal","finance","medical"],"locations":["San Francisco"],"founders":[{"name":"Rayan Krishnan","xHandle":null},{"name":"Langston Nashold","xHandle":null}],"teamSize":"1-10","funding":null,"raising":null,"founded":2024,"notable":"Vals Legal AI Report benchmarking legal AI vendors","entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/vals-ai"},{"slug":"veris-ai","name":"Veris AI","website":"https://veris.ai","tagline":"High-fidelity simulated environments to train enterprise AI agents","description":"Veris AI lets companies train and test AI agents in high-fidelity simulations of their real tools and workflows, so agents learn from experience rather than prompting alone. Early customers span financial services, enterprise productivity, and manufacturing.","domains":["enterprise","custom-environments","simulation"],"locations":["New York"],"founders":[{"name":"Mehdi Jamei","xHandle":null}],"teamSize":"1-10","funding":"Seed, $8.5M (Decibel and Acrew, June 2025)","raising":null,"founded":null,"notable":"Emerged from stealth June 2025","entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/veris-ai"},{"slug":"vetto-ai","name":"Vetto AI","website":"https://vetto.ai","tagline":"Code and computer-use environments from ex-DeepMind/Instagram founders","description":"Vetto AI builds coding and computer-use environments for frontier labs, with a founding team spanning San Francisco, São Paulo, and London including ex-DeepMind and ex-Instagram engineers.","domains":["coding","computer-use"],"locations":["San Francisco","São Paulo","London"],"founders":[{"name":"José André Nunes","xHandle":null},{"name":"Rodrigo Schmidt","xHandle":null},{"name":"Lucas Smaira","xHandle":null},{"name":"Gabriel Albuquerque","xHandle":null},{"name":"Roberta Antunes","xHandle":null}],"teamSize":"1-10","funding":null,"raising":null,"founded":null,"notable":null,"entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/vetto-ai"},{"slug":"vmax","name":"Vmax","website":"https://vmax.ai","tagline":"Converts proprietary data into RL environments","description":"Vmax builds RL/eval infrastructure that converts proprietary company data into training environments, including Unix/terminal task environments such as unix-ctf.","domains":["machine-learning","custom-environments"],"locations":["San Francisco","New York"],"founders":[{"name":"Matthew Sargent","xHandle":null},{"name":"Augustine Mavor-Parker","xHandle":null}],"teamSize":"1-10","funding":null,"raising":null,"founded":null,"notable":"unix-ctf environment","entityType":"Pure-play commercial","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/vmax"},{"slug":"genesis-ai","name":"Genesis AI","website":"https://genesis-ai.company","tagline":"Physics simulation engine and foundation model for robotics","description":"Genesis AI builds a universal robotics foundation model trained on synthetic data from its proprietary physics simulation engine, which generates high-fidelity training data up to 430,000x faster than real time. It emerged from stealth in July 2025 with a $105M seed round co-led by Eclipse Ventures and Khosla Ventures, and unveiled its GENE model line in 2026.","domains":["robotics","simulation","machine-learning"],"locations":["San Francisco","Paris"],"founders":[{"name":"Zhou Xian","xHandle":null},{"name":"Théophile Gervet","xHandle":null}],"teamSize":"26-50","funding":"Seed, $105M (Eclipse Ventures, Khosla Ventures, July 2025)","raising":null,"founded":2024,"notable":"Backers include Eric Schmidt and Bpifrance; open-sourced the Genesis physics simulator","entityType":"Simulator","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/genesis-ai"},{"slug":"duality-ai","name":"Duality AI","website":"https://duality.ai","tagline":"Falcon: reality-grade digital twin simulation for AI and robotics","description":"Duality AI builds Falcon, a digital twin simulation platform where teams train, test, and refine robots and physical AI systems in high-fidelity virtual environments that mirror real-world physics, sensors, and object dynamics. FalconEditor (built on Unreal Engine) simplifies digital twin creation, and FalconCloud runs simulations from the browser; the platform also generates synthetic training data for machine learning pipelines.","domains":["robotics","simulation"],"locations":["San Mateo"],"founders":[{"name":"Apurva Shah","xHandle":null},{"name":"Michael Taylor","xHandle":null}],"teamSize":"26-50","funding":null,"raising":null,"founded":2018,"notable":"Founding team from DreamWorks and Pixar; used in US government autonomy programs","entityType":"Simulator","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/duality-ai"},{"slug":"scaled-foundations","name":"Scaled Foundations","website":"https://scaledfoundations.ai","tagline":"GRID: a simulation-first platform for robot learning","description":"Scaled Foundations, founded by former Microsoft Research aerial robotics lead Ashish Kapoor, builds GRID, a platform that combines foundation models for perception, state estimation, safety, and control with simulation for rapid development of robot intelligence. GRID uses the AirGen simulator to generate training data and evaluate robotics policies, with an LLM-powered orchestration layer for natural interaction.","domains":["robotics","simulation","machine-learning"],"locations":["Seattle"],"founders":[{"name":"Ashish Kapoor","xHandle":null}],"teamSize":"1-10","funding":null,"raising":null,"founded":2023,"notable":"GRID-playground is open source; team behind Microsoft AirSim","entityType":"Simulator","research":null,"catalogCount":0,"canonicalUrl":"https://envindex.com/startup/scaled-foundations"}]}