Applied AI Scientist, Simulation & Reinforcement Learning
Locus Robotics
As Applied AI Scientist, you own and evolve our SimPy-based discrete-event simulator and use it to train and evaluate learning-based policies. You will drive sim-to-real alignment by comparing simulations to live warehouse data and expand the role toward reinforcement learning. You’ll work in a small, collaborative ADS team to bridge research and production, shaping a world-class orchestration framework for people and robots at global scale.
Pay / Benefits- bonus
- equity
- Own and develop the SimPy-based discrete-event simulator, including logic and design choices
- Improve test coverage, code quality, reproducibility, and developer tooling
- Build experiment infrastructure around the simulator for orchestration, study management, and automated runs
- Generate and maintain site data (distance files, configs) for warehouse simulation
- Support scenario testing and sensitivity analysis for order-engine and orchestration teams
- Define, measure and reduce the sim-to-real gap using real site data
- Run parameter optimisation and diagnose anomalies and non-determinism
- Model warehouse processes like congestion, picker behaviour and queueing at appropriate fidelity
- Communicate results clearly to engineering, product and operations stakeholders
- Turn the simulator into a RL environment with Gymnasium-style interfaces and scalable training runs
- Design, train and evaluate RL policies and benchmark against heuristics
- Prototype, evaluate and productionise learning components with ML, MLOps and OE teams
- Stay current with RL, multi-agent systems and simulation research and assess applicability
- Contribute to internal papers, patents, and publications when appropriate
- Hands-on experience building or validating simulators (preferably discrete-event like SimPy)
- Practical RL experience and simulation-based training
- Strong Python and software engineering fundamentals (testing, maintainable code)
- Rigorous experimental and statistical thinking (benchmarking, ablations, uncertainty)
- Experience in robotics, logistics, warehouse, or operations-heavy domains
- Publications in AI/ML, robotics or vision venues
- excellent communication with diverse audiences
- collaborative and cross-functional teamwork
- ability to explain uncertain results clearly to technical and non-technical audiences
- SimPy or similar discrete-event simulation
- Reinforcement learning and training simulation-based agents
- Gymnasium/Gym-style interfaces
Reference: WJ-747_31292041 · Posted 13 Nov 2026, 04:58 AM