Home›Companies›Epsilab

Long horizon RL environments from real-life data

Applied AI research lab focused on RL for frontier model training. Our research and evals cover real-world environment simulation, long-horizon workflows, and multi-player agents, etc.
Active Founders
Evan Hu
Evan Hu
Founder
Co-founder at Epsilab (F26). Previously co-founder and CTO at Shortbread (YC W23). Berkeley EECS '24
Founder at Epsilab - building RL environments for training agents on real workflows and ultra-long-horizon tasks.
Cofounder at Epsilab. Previously co-founded Shortbread (YC W23), generating $500K revenue and 2M episodes sold for AI-created comics. ML Researcher at MIT Media Lab.
Cofounder & CEO of Epsilab (F26). 2nd time YC founder - previously built Copper Health (W24).
Company Launches
Epsilab - RL environments for enterprise work
See original launch post

TLDR: Epsilab is an applied AI research lab building high-fidelity RL environments for long-horizon enterprise work. We partner with the world’s top professionals and work directly with frontier labs to improve model agentic capabilities in the real world.

We believe in human judgement and taste, and have seen that current frontier models have a long way to go when deployed in the real world. The widespread diffusion of AI depends on distilling the knowledge and expertise of the top performers in every field.

The problem

Today, low data quality is a pervasive, industry-wide problem. Widely used benchmarks like TerminalBench 4.0 and OSWorld contain many broken tasks with impossible criteria and unreliable graders susceptible to false positives and negatives. Designing RL environments and tasks that can realistically challenge frontier models without artificial handicaps increasingly requires not only deep domain expertise but also engineering, operational, and research excellence.

What makes us different

Seeing the gaps in existing benchmarks and RL environments firsthand, we founded Epsilab to focus on quality and realism of RL data. We ground our training gyms in real-world data and expert judgement from top industry professionals, so that our environments capture precisely how enterprise work actually gets done.

Maintaining this standard and level of fidelity is demanding and costly, and it is central to our thesis. Using our environments, frontier models and AI agents can hill climb on the right capabilities.

https://www.youtube.com/watch?v=LjYFbHWGPDQ

About us

  • Joanne is a 2x YC founder and previously built a healthcare AI startup Copper Health & worked in investment banking and PE
  • Fengjiao is also a 2x YC founder, did RL research at MIT Media Lab, worked as an ML researcher at Vimeo, and built an AI comics generation startup Shortbread
  • Evan previously was CTO and cofounder of Shortbread, and did research on LLMs at UC Berkeley
  • Matthew worked as a quant infra engineer at Tower Research Capital, an HFT firm

Reach out

Jobs at Epsilab
San Francisco, CA, US / Remote (US)
$105K - $145K
0.50% - 1.00%
1+ years
San Francisco, CA, US
$125K - $300K
0.75% - 2.00%
3+ years
Epsilab
Founded:2026
Batch:Fall 2026
Team Size:4
Status:
Active
Location:San Francisco
Primary Partner:Pete Koomen