hud
Platform for building RL environments and evals
Y CombinatorWinter 2025ActiveB2BEngineering, Product and DesignArtificial IntelligenceReinforcement Learning
What they do
Problem
AI labs lack reliable, comprehensive ways to evaluate and train computer-use agents across diverse real-world tasks.
Solution
HUD provides a platform and framework for building agentic evaluations and RL environments to test and improve web-browsing computer-use agents at scale.
Who it's for
Frontier AI labs and teams building computer-use agents use it to evaluate and train their agents.
Founders
Jay Ram
Founder
CEO @ HUD prev: consumer apps, ml research, quant research
Lorenss Martinsons
Founder
CPO @ hud interested in natural intelligence, flowers, and the human condition
Parth Patel
Founder
CTO @ HUD. [a-z]*
Open roles 5
Company timeline
- Founded2025
- YC batchWinter 2025
- StatusActive
- Team size15 people
- LocationSan Francisco, CA, USA
Similar startups
Envariant
Interpretability and reasoning infra for foundation models.
Coval
Simulation & Evaluation that scales voice and chat AI agents
Playgent
Sandboxes for AI agents
MangoDesk
Long-horizon RL environments
camelAI
AI software engineer that lives on its own computer
Baserun
Observability and evaluation platform for LLM apps.