Website:
lat.com
Job details:
We're Hiring:
Reinforcement Learning Engineer
Full time | Core team
LAT Aerospace | Perception & Autonomy
At LAT Aerospace, we’re building India’s first clean-sheet hybrid-electric STOL aircraft and a next-generation autonomy stack powering aircraft for complex missions. Our team is kicking off a new chapter — developing the full flight-control, obstacle-avoidance, GNSS-denied navigation, and swarm-coordination layers that will define LAT’s autonomy architecture.
As an RL Engineer, you will design, train, and deploy policy learning systems that optimize guidance decisions for multi-agent swarms. Your algorithms will directly shape LAT’s next-generation autonomous behaviors.
What You’ll Do
- Develop reinforcement learning and policy-learning algorithms that improve swarm guidance, collision avoidance, task allocation, and distributed decision-making.
- Integrate RL with classical control and planning — combining learned value functions, cost shaping, and residual control to boost agility, robustness, and safety.
- Build scalable training pipelines for multi-agent RL using domain randomization, curriculum learning, simulation rollouts, and extensive scenario generation.
- Own sim-to-real transfer, ensuring policies trained in simulation generalize to real-world UAV dynamics, uncertainties, and edge-case environments.
- Design multi-agent coordination behaviors such as formation flight, coverage, pursuit/avoidance, collaborative mapping, and decentralized cooperation under minimal communication.
- Run frequent field tests to evaluate learned policies on actual UAVs, gather flight data, and iterate rapidly.
- Develop evaluation frameworks and debugging tools to diagnose RL failures, mode collapse, instability, or unsafe behaviors.
What We're Looking For
- Strong fundamentals in reinforcement learning, policy optimization (PPO, SAC, TD3), or multi-agent RL.
- Experience with robotics autonomy, motion planning, or control systems.
- Proficiency in Python with RL libraries (PyTorch, JAX, RLlib, Stable Baselines, CleanRL, etc.).
- Hands-on experience with robotics simulation environments — Isaac Lab, Gazebo, MuJoCo, PyBullet, or custom simulators.
- Comfort with integrating RL modules into larger autonomy frameworks and evaluating them on real systems.
Why LAT?
- Core-team role building the autonomy brain of India’s most ambitious aerospace startup.
- Ownership, speed, and the opportunity to define a new class of systems from scratch.
Click on Apply to know more.