Vmax
Vmax is building RL environments for long-horizon tasks that capture and simulate the ineffable.
South Park Commons2025SeedGeneral Software
What they do
Problem
Reinforcement learning researchers and developers lack realistic environments for training agents on long-horizon tasks that involve hard-to-formalize real-world dynamics.
Solution
Vmax builds reinforcement learning environments that simulate long-horizon tasks and capture otherwise ineffable aspects of real-world behavior.
Who it's for
Reinforcement learning researchers and developers building and training RL agents.
Founders
Matthew Sargent
PhD in RL, building new ways of using RL + LLMs
Augustine Mavor-Parker
RL PhD at UCL, working on LLMs and reasoning at vmax, prev: AI safety at Redwood research, NeuroAI at CSHL, AI for genomics at lllumina
Open roles 3
Company timeline
- Founded2025
- StatusSeed
- LocationSan Francisco
Similar startups
Symplex
Applied Reinforcement Learning Research Laboratory
Preference Models
Preference Model is building reinforcement learning environments.
Fulcrum
The agentic debugger for AI systems
MangoDesk
Long-horizon RL environments
Polymath
Simulation environments to train & evaluate long-horizon AI agents
Vibrant Labs
RL environments for long horizon AI Agents