Research Engineer
Atul Saxena
I build the environments models are trained in, and the tools that show what they learned.
- RL Systems
- Interpretability
- ML Infrastructure
Research Engineer
Atul Saxena
I build the environments models are trained in, and the tools that show what they learned.
Photorealistic drone simulation for vision-language-action research
A pip-installable drone environment that combines 6-DOF quadrotor dynamics with photorealistic Gaussian-Splatting rendering, and the data tooling to train and evaluate VLA policies on it.
VLA-AN identifies the visual sim-to-real gap as a primary blocker for drone VLA systems, and RaceVLA shows aerial policies work but degrade on safety and generalisation. Neither problem is addressable without a simulator that renders convincingly and a data layer that speaks the formats the field already uses.
status: Active · research infrastructure