Reinforcement Learning for High-Fidelity Flight Simulation (Research/Engineering)

📍 Bucharest / Remote (EU) | Full-time | part-timePosition: Reinforcement Learning Engineer — Flight SimulationLocation: RemoteEmployment type: Full-time / Part-timeReports to: Director of Autonomy Warfare SystemsWe are building robust, safe, and highly realistic flight autonomy systems by training reinforcement learning (RL) agents in high-fidelity simulators and transferring learned policies to physical aircraft under human supervision. You will design simulation environments, implement RL training pipelines, and lead sim-to-real validation (SIL / HIL / shadow testing) using JSBSim, NVIDIA Omniverse / Isaac Sim, and modern RL toolchains.What you’ll doDesign, implement and validate flight simulation environments and mission scenarios with realistic flight dynamics, sensors, and actuator models (using JSBSim, FlightGear/X-Plane backends as applicable).Develop and maintain RL training pipelines (single-agent and multi-agent) using state-of-the-art algorithms (PPO, SAC, MAPPO, etc.), distributed/external rollout collectors, experiment tracking and reproducible CI.Implement domain randomization, system identification, and domain adaptation techniques to maximize sim-to-real transferability.Integrate GPU-accelerated simulation and sensor synthesis (NVIDIA Isaac Sim / Omniverse) for large-scale parallel training and synthetic perception data generation.Build safe hybrid control stacks: stable low-level controllers (PID/LQR) combined with high-level RL policies; implement runtime safety supervisors, geofencing, and fallback behaviors.Lead validation workflows: Software-in-the-Loop (SIL), Hardware-in-the-Loop (HIL), Flight-in-the-Loop (FIL), shadow mode tests and progressive flight testing under human oversight.Collaborate with systems engineers, avionics, and safety engineers to define safety cases, test plans, and verification/validation artefacts.Produce clear documentation, reproducible experiments, and publishable technical reports or internal design reviews.What you bring BSc or MSc in Robotics, Control, Aerospace Engineering, Computer Science, Artificial Intelligence or equivalent practical experience.Experience developing RL solutions or advanced control for physical systems (aircraft, UAVs, robotics).Strong understanding of reinforcement learning fundamentals: MDPs, policy/value functions, actor-critic methods, on/off-policy tradeoffs, exploration/exploitation, reward shaping, and sample efficiency techniques.Practical experience with at least one major RL framework (Stable-Baselines3, RLlib, Acme, CleanRL, or custom PyTorch/TensorFlow implementations).Experience working with flight dynamics models (JSBSim, FlightGear, X-Plane) or direct aerospace dynamics modelling experience.Hands-on experience with simulation integration and environment wrapping (OpenAI Gym / Gymnasium style APIs).Experience with GPU-accelerated simulation or physics engines (NVIDIA Isaac Gym / Isaac Sim / Omniverse, or equivalent), including running many parallel environments for RL training.Solid software engineering skills: Python (primary), C++ (nice to have), testable, modular code, unit/integration tests, CI pipelines.Experience with experiment logging and reproducibility: Weights & Biases, TensorBoard, MLFlow or similar.Clear understanding of safety, ethics and regulatory constraints when developing autonomy for airborne systems.Bonus points- Prior work on sim-to-real transfer for aerial vehicles or other inertia-dominated systems.- Background with system identification, Kalman/EKF/UKF state estimation, sensor fusion and perception pipelines (camera, IMU, LiDAR).- Experience with autopilot stacks (PX4, ArduPilot) and SITL/HITL testing frameworks.- Experience with multi-agent RL and self-play methods for adversarial/cooperative scenarios (MAPPO, centralized critic architectures).- Familiarity with radar/IR/sensor models, electronic warfare modelling, or other domain-specific sensor suites (civilian/non-lethal contexts).- Experience in structured safety engineering: writing safety cases, fault tree analysis (FTA), or failure modes and effects analysis (FMEA).Soft Skills & Competencies:- Strong problem solving and principled engineering judgment; ability to decompose large sim-to-real problems into incremental experiments.- Excellent communication: translate technical tradeoffs into concise decisions for domain experts and management.- Collaborative: cross-disciplinary teamwork with avionics, test pilots, legal & compliance.- Attention to documentation, reproducibility and experimental rigor.- Commitment to ethical development and safety-first deployment.What we offerOpportunity to work on cutting-edge autonomy research and real-world flight testing (civilian / research / commercial applications).The chance to be part of a unique, ground-breaking project in Romania, working with advanced technologies and dedicated tools.Collaboration with cross-functional teams (systems engineering, test pilots, ML researchers).Flexible working arrangements.