RL Playground

Pure JAX/Flax implementations of the RL algorithm ladder — DQN, A2C, and DDPG — with no Stable-Baselines or RLlib dependencies. Every update step is a JIT-compiled pure function. The pipeline includes multi-modal observation engines, Gymnasium and MuJoCo environments, and an ONNX export path targeting NVIDIA Jetson Orin and Thor for real-world robotics edge deployment via TensorRT.

Capabilities applied

Technology