Service
Active inference & RL
Free-energy formulations, exploration-exploitation, learned policies.
- Typical engagement
- 12-20 weeks
- Capabilities
- free-energy formulationsPPO / SAC variantslearned planning
Active inference and RL applied to real systems — not toy gridworlds.
Active inference and reinforcement learning, applied to systems that have to make decisions under partial information. Free-energy formulations where they fit, PPO / SAC where they don’t.
We start from the task, not the algorithm. What does the agent know, what does it want, what does it pay to find out? Most of the wins live in the reward shaping and the state representation, not the optimizer.
We don’t pretend a learned policy is a substitute for a tested one. The policy goes into a system you can still reason about.