Job Description

1) RL Control for High-Speed ​​Drones and Robotics - Design RL-based control algorithms for high-speed drones and other robotic platforms - Continuous control policy learning and iterative improvement utilizing cutting-edge RL techniques (e.g., PPO, SAC, TD3) - Define safety constraints, guardrails, and fallback (alternative) controllers to ensure reliability in real-world environments - Explore RL-based approaches to improve rover navigation and energy efficiency 2) Simulation, HITL, and Robustness Testing - Build and maintain a simulation environment (Gazebo, Isaac, UnrealEngine, or custom) that reflects real-world hardware behavior - Design RL scenarios, curriculum strategies, and sim-to-real pipelines for robust policy transfer - Perform HITL and adversarial testing to validate under edge cases and performance degradation - Standardize simulation outputs (rollouts, rewards, metrics) for reproducible experiments 3) RL Data Workflow (Logs, Sensors, Videos, Datasets) - Flight logs, Own RL experimentation data workflow across sensor data, video, and simulation rollouts. - Define naming conventions, metadata schemas, and storage layouts to ensure easy discovery and consistent use of datasets. - Curate and version datasets for RL training/evaluation (Train/Val/Test splits, dataset versioning). - Design a dataset strategy that carefully considers coverage, bias, and failure modes. 4) ML for Perception/Tracking (if required for RL/Autonomous) - Build and maintain ML pipelines for computer vision tasks (detection, tracking, etc.) that support Autonomous/RL. - Integrate perception results into RL training and evaluation loops, as needed. 5) Edge Deployment and Model Optimization - Deploy RL policies and ML models to edge hardware (e.g., Jetson/embedded GPUs). - Optimize models considering on-device constraints: distillation, quantization, pruning, and runtime performance tuning. - Ensure consistency in evaluation/validation between offline experiments and on-device operations. 6) Experiment Management and Cross-Functional Collaboration - Define an experiment management system: run ID, configuration, metrics, artifact storage, and dashboards - Collaborate with full-stack engineers to design and build a storage system, database schema, and internal tools for log exploration and experiment tracking - Collaborate with autonomous driving engineers to drive secure control interfaces, shared simulation assets, and full autonomous stack integration - Collaborate with LLM engineers when RL policies need to be exposed as higher-level "skills" or tools

Qualifications

- 5+ years of experience applying RL/ML in a production or research-to-deployment environment - Excellent Python skills and hands-on experience with PyTorch or TensorFlow - Hands-on experience with RL algorithms for continuous control (PPO, SAC, TD3, etc.) and large-scale training - Experience designing datasets and data pipelines (logging, metadata, versioning, Train/Val/Test splits) - Experience deploying ML/RL policies and optimizing performance on edge hardware (Jetson, embedded GPUs, etc.)

Preferred

- Background in control theory and motion planning (MPC, trajectory optimization, sampling-based planning, etc.) - Understanding/experience with Ardupilot, PX4, or industrial robot arms - Experience with robotics frameworks (ROS/ROS2) and simulators (Gazebo, Isaac, UnrealEngine, etc.) - Excellent communication skills (fluent English preferred)

Preferred Visas

Cover Letter

Optional