Directory

Coding Atlas

Who sells coding data for AI, by category. 51 companies across 5 categories, researched by AI Circle. Explore them on the interactive map above, or browse the full list below.

SFT & Code Generation19

Expert-written code, prompt-solution pairs, multi-turn coding conversations, SFT datasets, and repository-level tasks from vetted engineers.

  • AfterQuerySan Francisco, CA

    Expert human data company producing coding and reasoning data, preference data, and custom evaluations for frontier labs.

  • Beyond Labs

    Human data for code generation and evaluation. Limited public information.

  • DatacurveSan Francisco, CA · Founded 2024

    Coding data company that sources expert software engineers through a bounty-style platform to produce SFT, preference, evaluation, and agentic coding data for frontier labs.

  • EPAMNewtown, PA · Founded 1993

    Global software engineering and digital services company with a large engineering workforce, providing AI engineering work that includes code generation and evaluation data for model builders.

  • G2iFounded 2016

    Engineering talent network that supplies senior developers for AI training work, including code generation and code evaluation.

  • Handshake AISan Francisco, CA · Founded 2014

    Human data arm of the Handshake student and early-career network, recruiting CS students and graduates for coding SFT and evaluation work.

  • Invisible TechnologiesNew York, NY · Founded 2015

    AI training and operations company with an expert workforce that produces SFT and RLHF data, including coding, for frontier labs.

  • MercorSan Francisco, CA · Founded 2023

    AI-driven expert marketplace that recruits and vets software engineers and domain experts for labs, covering code SFT, preference data, and eval creation.

  • Poindexter LabsLondon, UK · Founded 2025

    Produces software engineering data for frontier AI labs alongside its STEM, translation and safety work, using expert contributors reviewed with an academic peer-review model. It reports $9M ARR and 99.95% accuracy as measured by its lab customers, and has published at ACL and EMNLP, with two papers at NeurIPS and mini-benchmarks on Hugging Face.

  • ReveloMiami, FL · Founded 2015

    Data lab specializing in code for frontier AI. It says engineers from its network of 400,000+ produce expert trajectories for SFT, preference data for RLHF, expert-graded evals and benchmarks, and RL environments for agents across terminal coding, frontend and web apps, game development, cybersecurity and other work done through code.

  • Scale AISan Francisco, CA · Founded 2016

    Largest human data vendor, running coding SFT and RLHF programs through its expert contributor network alongside evaluation and leaderboard work.

  • Snorkel AIRedwood City, CA · Founded 2019

    Data development company that pairs programmatic labeling with expert networks to deliver coding datasets, RL environments, and evaluation for labs.

  • SuperAnnotateSan Francisco, CA · Founded 2018

    Annotation platform with a managed expert workforce, used for coding SFT, preference data, evaluation, and agent trajectory work.

  • Surge AISan Francisco, CA · Founded 2020

    Premium human data company supplying expert-written and expert-rated data, including coding SFT and preference data, to frontier labs.

  • TbrainFlorida, US / Hanoi, Vietnam · Founded 2025

    Runs human data programs for frontier AI teams across coding, agentic workflows, evaluation and post-training, alongside its physical AI capture work. It covers SFT and code generation, RLHF and code evaluation, benchmarks and evals, including Terminal Bench work, and agentic coding environments, using expert contributors and LLM-as-a-judge checks before final human review.

  • Tiki AI

    Coding data for AI training. Limited public information.

  • TuringPalo Alto, CA · Founded 2018

    Developer network turned AI data company, supplying vetted engineers for coding SFT, RLHF, reasoning data, evals, and secure-code work for model builders.

  • Vetto AI

    Post-training data company that pays domain experts per task to build RL environments, verifier-graded benchmarks and human evaluations. Its coding work includes the SlopCode Vetted and ProgramBench Vetted leaderboards, and it reports expert-written SFT and code preference data.

  • Vraify

    Coding data for model training. Limited public information.

RLHF & Code Evaluation18

Pairwise preferences, rubric grading, code review and critique, reward-model data, and human evaluation of model-written code.

  • AfterQuerySan Francisco, CA

    Expert human data company producing coding and reasoning data, preference data, and custom evaluations for frontier labs.

  • Beyond Labs

    Human data for code generation and evaluation. Limited public information.

  • DatacurveSan Francisco, CA · Founded 2024

    Coding data company that sources expert software engineers through a bounty-style platform to produce SFT, preference, evaluation, and agentic coding data for frontier labs.

  • EPAMNewtown, PA · Founded 1993

    Global software engineering and digital services company with a large engineering workforce, providing AI engineering work that includes code generation and evaluation data for model builders.

  • G2iFounded 2016

    Engineering talent network that supplies senior developers for AI training work, including code generation and code evaluation.

  • Guilded AI

    Expert data foundry that sources training and evaluation data from specialist contributors, with recurring evaluations, documented data provenance, and long-term incentives for the experts who produce the data.

  • Handshake AISan Francisco, CA · Founded 2014

    Human data arm of the Handshake student and early-career network, recruiting CS students and graduates for coding SFT and evaluation work.

  • MercorSan Francisco, CA · Founded 2023

    AI-driven expert marketplace that recruits and vets software engineers and domain experts for labs, covering code SFT, preference data, and eval creation.

  • ProlificLondon, UK · Founded 2014

    Participant platform with verified, demographically diverse contributors, including developers, for human evaluation and preference data.

  • ReveloMiami, FL · Founded 2015

    Data lab specializing in code for frontier AI. It says engineers from its network of 400,000+ produce expert trajectories for SFT, preference data for RLHF, expert-graded evals and benchmarks, and RL environments for agents across terminal coding, frontend and web apps, game development, cybersecurity and other work done through code.

  • Scale AISan Francisco, CA · Founded 2016

    Largest human data vendor, running coding SFT and RLHF programs through its expert contributor network alongside evaluation and leaderboard work.

  • Snorkel AIRedwood City, CA · Founded 2019

    Data development company that pairs programmatic labeling with expert networks to deliver coding datasets, RL environments, and evaluation for labs.

  • SuperAnnotateSan Francisco, CA · Founded 2018

    Annotation platform with a managed expert workforce, used for coding SFT, preference data, evaluation, and agent trajectory work.

  • Surge AISan Francisco, CA · Founded 2020

    Premium human data company supplying expert-written and expert-rated data, including coding SFT and preference data, to frontier labs.

  • TbrainFlorida, US / Hanoi, Vietnam · Founded 2025

    Runs human data programs for frontier AI teams across coding, agentic workflows, evaluation and post-training, alongside its physical AI capture work. It covers SFT and code generation, RLHF and code evaluation, benchmarks and evals, including Terminal Bench work, and agentic coding environments, using expert contributors and LLM-as-a-judge checks before final human review.

  • Tiki AI

    Coding data for AI training. Limited public information.

  • TuringPalo Alto, CA · Founded 2018

    Developer network turned AI data company, supplying vetted engineers for coding SFT, RLHF, reasoning data, evals, and secure-code work for model builders.

  • Vetto AI

    Post-training data company that pays domain experts per task to build RL environments, verifier-graded benchmarks and human evaluations. Its coding work includes the SlopCode Vetted and ProgramBench Vetted leaderboards, and it reports expert-written SFT and code preference data.

Benchmarks & Evals21

Custom and private coding benchmarks, SWE-style task sets, eval harnesses, leaderboards, and evaluation platforms.

  • AfterQuerySan Francisco, CA

    Expert human data company producing coding and reasoning data, preference data, and custom evaluations for frontier labs.

  • BenchFlow

    Infrastructure for running AI benchmarks and agent evaluations in reproducible environments.

  • BraintrustSan Francisco, CA · Founded 2023

    Evaluation and observability platform that teams use to run evals, compare prompts and models, and monitor AI apps in production.

  • DatacurveSan Francisco, CA · Founded 2024

    Coding data company that sources expert software engineers through a bounty-style platform to produce SFT, preference, evaluation, and agentic coding data for frontier labs.

  • Guilded AI

    Expert data foundry that sources training and evaluation data from specialist contributors, with recurring evaluations, documented data provenance, and long-term incentives for the experts who produce the data.

  • HashlistHelsinki, Finland

    Helsinki-based engineering data company (Gigarator Oy) that builds custom evals, datasets and RL environments for embedded engineering. It publishes ConstraintBench, a benchmark of chip design, RTL, CAD and embedded coding tasks written by domain engineers and graded by scripts against each design's real constraints.

  • idler

    Benchmarks, environments, and security testing for AI agents. Limited public information.

  • MercorSan Francisco, CA · Founded 2023

    AI-driven expert marketplace that recruits and vets software engineers and domain experts for labs, covering code SFT, preference data, and eval creation.

  • Patronus AISan Francisco, CA · Founded 2023

    Evaluation platform for LLMs and agents covering automated evaluators, benchmarks, safety testing, and agent environments.

  • Perle AI

    Expert-in-the-loop data annotation and evaluation for AI models.

  • Poindexter LabsLondon, UK · Founded 2025

    Produces software engineering data for frontier AI labs alongside its STEM, translation and safety work, using expert contributors reviewed with an academic peer-review model. It reports $9M ARR and 99.95% accuracy as measured by its lab customers, and has published at ACL and EMNLP, with two papers at NeurIPS and mini-benchmarks on Hugging Face.

  • ProlificLondon, UK · Founded 2014

    Participant platform with verified, demographically diverse contributors, including developers, for human evaluation and preference data.

  • Proximal

    Evaluations and RL environments for AI agents. Limited public information.

  • ReveloMiami, FL · Founded 2015

    Data lab specializing in code for frontier AI. It says engineers from its network of 400,000+ produce expert trajectories for SFT, preference data for RLHF, expert-graded evals and benchmarks, and RL environments for agents across terminal coding, frontend and web apps, game development, cybersecurity and other work done through code.

  • Specific Labs

    Evaluations and environments for coding agents. Limited public information.

  • SuperAnnotateSan Francisco, CA · Founded 2018

    Annotation platform with a managed expert workforce, used for coding SFT, preference data, evaluation, and agent trajectory work.

  • SWEbench.ai (Bhavitech)

    SWE-bench-style software engineering task data and evaluation. Limited public information.

  • TbrainFlorida, US / Hanoi, Vietnam · Founded 2025

    Runs human data programs for frontier AI teams across coding, agentic workflows, evaluation and post-training, alongside its physical AI capture work. It covers SFT and code generation, RLHF and code evaluation, benchmarks and evals, including Terminal Bench work, and agentic coding environments, using expert contributors and LLM-as-a-judge checks before final human review.

  • TuringPalo Alto, CA · Founded 2018

    Developer network turned AI data company, supplying vetted engineers for coding SFT, RLHF, reasoning data, evals, and secure-code work for model builders.

  • Vals AISan Francisco, CA · Founded 2023

    Independent evaluation company publishing industry benchmarks and running custom evals of models and agents, including coding tasks.

  • Vetto AI

    Post-training data company that pays domain experts per task to build RL environments, verifier-graded benchmarks and human evaluations. Its coding work includes the SlopCode Vetted and ProgramBench Vetted leaderboards, and it reports expert-written SFT and code preference data.

Agentic Coding & Environments27

RL environments, verifiable tasks, agent trajectories, sandboxed repos and tools, and computer-use or terminal environments.

  • Align AI

    Agent environments and training data. Limited public information.

  • BenchFlow

    Infrastructure for running AI benchmarks and agent evaluations in reproducible environments.

  • Bespoke LabsFounded 2024

    Data curation and synthetic data tooling, including open reasoning and coding datasets and RL environments for agents.

  • Cartpole

    RL environments for AI agents. Limited public information.

  • Chakra Labs (Dojo)

    Dojo builds environments and task data for training and evaluating AI agents.

  • DatacurveSan Francisco, CA · Founded 2024

    Coding data company that sources expert software engineers through a bounty-style platform to produce SFT, preference, evaluation, and agentic coding data for frontier labs.

  • Deeptune

    RL environments and training data for AI agents. Limited public information.

  • Fleet

    RL environments for training AI agents. Limited public information.

  • Frist

    Early-stage agentic coding environments company. Limited public information; no public website yet.

  • Guilded AI

    Expert data foundry that sources training and evaluation data from specialist contributors, with recurring evaluations, documented data provenance, and long-term incentives for the experts who produce the data.

  • HashlistHelsinki, Finland

    Helsinki-based engineering data company (Gigarator Oy) that builds custom evals, datasets and RL environments for embedded engineering. It publishes ConstraintBench, a benchmark of chip design, RTL, CAD and embedded coding tasks written by domain engineers and graded by scripts against each design's real constraints.

  • HUDSan Francisco, CA

    Platform for building and running RL environments and evaluations for computer-use and coding agents.

  • idler

    Benchmarks, environments, and security testing for AI agents. Limited public information.

  • Matrices

    Agent environments and training data. Limited public information.

  • MechanizeSan Francisco, CA · Founded 2025

    Builds RL environments for software engineering and other white-collar work, aimed at training agents on realistic long-horizon tasks.

  • Polymath

    Environments and task data for AI agents. Limited public information.

  • Preference Model

    RL environments and preference data for AI agents. Limited public information.

  • Prime IntellectSan Francisco, CA · Founded 2024

    Decentralized training company that runs an open hub of RL environments and verifiers alongside distributed compute.

  • Proximal

    Evaluations and RL environments for AI agents. Limited public information.

  • Refresh

    Coding environments and security-focused tasks for AI agents. Limited public information.

  • ReveloMiami, FL · Founded 2015

    Data lab specializing in code for frontier AI. It says engineers from its network of 400,000+ produce expert trajectories for SFT, preference data for RLHF, expert-graded evals and benchmarks, and RL environments for agents across terminal coding, frontend and web apps, game development, cybersecurity and other work done through code.

  • Snorkel AIRedwood City, CA · Founded 2019

    Data development company that pairs programmatic labeling with expert networks to deliver coding datasets, RL environments, and evaluation for labs.

  • Specific Labs

    Evaluations and environments for coding agents. Limited public information.

  • SuperAnnotateSan Francisco, CA · Founded 2018

    Annotation platform with a managed expert workforce, used for coding SFT, preference data, evaluation, and agent trajectory work.

  • TbrainFlorida, US / Hanoi, Vietnam · Founded 2025

    Runs human data programs for frontier AI teams across coding, agentic workflows, evaluation and post-training, alongside its physical AI capture work. It covers SFT and code generation, RLHF and code evaluation, benchmarks and evals, including Terminal Bench work, and agentic coding environments, using expert contributors and LLM-as-a-judge checks before final human review.

  • Vetto AI

    Post-training data company that pays domain experts per task to build RL environments, verifier-graded benchmarks and human evaluations. Its coding work includes the SlopCode Vetted and ProgramBench Vetted leaderboards, and it reports expert-written SFT and code preference data.

  • Vmax

    Agent environments and training data. Limited public information.

Security & Vulnerability11

Vulnerability discovery data, secure-coding evals, red-teaming, exploit and CTF-style tasks, and agent safety testing.

  • Andon LabsSan Francisco, CA · Founded 2023

    AI safety company that stress-tests autonomous agents in long-running real-world and simulated deployments, publishing evals such as Vending-Bench.

  • Deaimer

    Security and vulnerability data for AI. Limited public information.

  • Fabraix

    Security testing for AI systems and agents. Limited public information.

  • idler

    Benchmarks, environments, and security testing for AI agents. Limited public information.

  • Noumad

    Security and vulnerability data for AI. Limited public information.

  • Patronus AISan Francisco, CA · Founded 2023

    Evaluation platform for LLMs and agents covering automated evaluators, benchmarks, safety testing, and agent environments.

  • Refresh

    Coding environments and security-focused tasks for AI agents. Limited public information.

  • ReveloMiami, FL · Founded 2015

    Data lab specializing in code for frontier AI. It says engineers from its network of 400,000+ produce expert trajectories for SFT, preference data for RLHF, expert-graded evals and benchmarks, and RL environments for agents across terminal coding, frontend and web apps, game development, cybersecurity and other work done through code.

  • Trajectory Labs

    Security and vulnerability data for AI. Limited public information.

  • TuringPalo Alto, CA · Founded 2018

    Developer network turned AI data company, supplying vetted engineers for coding SFT, RLHF, reasoning data, evals, and secure-code work for model builders.

  • Vals AISan Francisco, CA · Founded 2023

    Independent evaluation company publishing industry benchmarks and running custom evals of models and agents, including coding tasks.