Featured Project
Meta-Learning RL System for Non-Stationary Environments
0.73-0.75 success vs. 0.46 baseline
Engineered a dual-agent PPO system with neuromodulation to adapt under shifting conditions while reducing catastrophic forgetting across task changes.
open project page
offline
static
live
- Designed coordinated policy and adaptation loops using PyTorch and PPO.
- Mitigated catastrophic forgetting in non-stationary training regimes.
- Repeatedly recovered 80-95% performance after distribution shifts.
PyTorch
PPO
Python
Meta-Learning
