Surge AI
Surge AI’s mission is to raise AGI with the richness of human intelligence—curious, witty, creative, and wise. It builds expert-produced training data, evaluations, reinforcement-learning environments, and human-feedback programs for frontier and enterprise AI systems.
At a Glance
- Frontier AI research labs
- Enterprise software and AI teams
- Technology companies building or evaluating LLMs and agents
- Healthcare, finance, tax, accounting, legal, engineering, manufacturing, insurance, logistics, HR, and public-sector organizations
- +1 more
AI Tools by Surge AI
(1)
Surge AI
Human Data for AI Training
Discussions
No discussions yet
Be the first to start a discussion about Surge AI
Latest News
Six Surge benchmarks joined the Fireworks Specialized Intelligence Index, covering professional work and software engineering.
Surge released DAYJOB: Tax, a 50-environment benchmark for AI agents performing complex professional tax work.
Surge released sudo L7, a 60-task staff-level software-engineering benchmark focused on code, risk, and professional judgment.
Surge released GDP.xlsx, a 70-task benchmark testing agents on real-world professional spreadsheets across 12 domains.
Products & Services
Custom evaluation, improvement, and training programs: workflow-specific benchmarks, launch validation, production monitoring, diagnosis, expert feedback, supervised fine-tuning, preference optimization, and custom RL environments.
Pre-built training datasets, evaluations, and RL environments for coding agents, enterprise agents, multimodal reasoning, and expert reasoning; access is requested through Surge’s catalog and Trusted Program.
Rich, complex environments and verifiers for training agentic models on long-horizon tool use, enterprise workflows, coding, and professional work.
Expert-designed scoring systems and verifiers that measure nuanced correctness, judgment, safety, and professional quality.
Market Position
Surge positions itself as an expert-data and evaluation company for frontier AI, differentiated from generic crowdsourcing by domain-qualified contributors, expert-authored rubrics, realistic RL environments, and long-horizon evaluations. Its closest competitive set includes Scale AI, Labelbox, Appen, Sama, Mercor, and Invisible; Surge emphasizes quality, judgment, confidentiality, and post-training outcomes rather than annotation volume alone.
Leadership
Founders
Edwin Chen
Founder and CEO. Chen studied mathematics, computer science, and linguistics at MIT and previously worked in machine learning/data science at Google, Facebook/Meta, Twitter, and Microsoft Research; Crunchbase identifies him as Meta’s former AI lead.
Executive Team
Edwin Chen
Founder and Chief Executive Officer
MIT-trained mathematician, computer scientist, and linguist; former machine-learning/data-science leader at Google, Facebook/Meta, Twitter, and Microsoft Research.
Founding Story
Chen started Surge in 2020 after becoming dissatisfied with the quality and context-insensitivity of crowdsourced data labeling. The initial vision was to build a higher-quality human-data infrastructure for AI, using skilled annotators and strong quality controls rather than treating labeling as undifferentiated volume; the company was bootstrapped and profitable from the outset.
Business Model
Revenue Model
B2B contracts for custom training data, human feedback, evaluations, benchmarks, RL environments, and enterprise AI programs, supplemented by licensing/access to off-the-shelf datasets. The company also offers a Trusted Program in which labs can train on a dataset and pay if it improves agreed metrics.
Target Markets
- Frontier AI research labs
- Enterprise software and AI teams
- Technology companies building or evaluating LLMs and agents
- Healthcare, finance, tax, accounting, legal, engineering, manufacturing, insurance, logistics, HR, and public-sector organizations
- Academic and alignment/safety research groups
- Training and post-training frontier language models
- RL for coding agents and terminal-use agents
- Enterprise tool-use and long-horizon workflow agents
- Model selection, launch-readiness evaluation, and production monitoring
- Professional-domain AI in tax, finance, healthcare, law, engineering, and data science
- Multimodal document, spreadsheet, chart, image, audio, and video reasoning
- OpenAI
- Anthropic
- Meta