Reinforcement Learning Engineer (m/w/d)

Project type: Freelance / T&M

**Start:** Mid-August 2026

**Workload:** approx. 200 hours until end of September 2026, with option to extend until March 2027 (T&M)

**Location:** Remote — delivery must be performed from within Germany

**Project language:** English or German

Project Context

We are looking for an experienced RL engineer specialized in deep reinforcement learning for continuous control problems. Further project and client details will be shared under NDA with shortlisted candidates.

Required Profile

**Must-have:**

• Demonstrable hands-on experience with deep RL for continuous action spaces (not just discrete-action RL such as classic DQN/Atari)
• Solid working experience with Stable Baselines3, PPO and/or SAC
• Experience with POMDP techniques: recurrent policies, frame stacking, belief-state modeling, or asymmetric actor-critic
• Experience with domain randomization and/or curriculum learning for sim-to-real robustness
• Clean experiment infrastructure practices: hyperparameter sweeps, seed management, reproducible evaluation
• Ability to honestly and quantitatively benchmark RL results against classical control baselines
• Production-grade Python

**Nice-to-have:**

• Background in classical control theory / optimal control (MPC, Kalman filtering)
• Experience with reward design for complex, multi-objective targets
• Experience working with digital twins / simulation environments as training grounds

Application

Please submit your profile along with concrete project references demonstrating RL experience (technology stack, type of action space, your role in the project). Due to a tight client timeline, profiles will be reviewed on a rolling basis.