Reinforcement Learning Engineer, Game Automation and Balance Optimization

This freelance role designs, develops, and deploys AI systems that use reinforcement learning for automated game play and game-balance optimization. Responsibilities include building automated play agents for QA and test automation, player bots, and optimized navigation AI. The role also applies self-play and reward design to balance optimization, improves NPC and agent behavior models, and develops adaptive AI based on player skill. Other work includes building reinforcement-learning models from game-play data, including imitation learning and simulated environments, implementing and optimizing reinforcement-learning algorithms in cloud and distributed-processing environments, and creating workflows with game designers and QA teams. Technical blogging is also part of the role.

Required qualifications are knowledge of reinforcement-learning and machine-learning theory and implementation; development experience with Python and C++ or C#; experience using deep-learning frameworks such as TensorFlow, PyTorch, or JAX; experience developing models with reinforcement-learning libraries including Ray RLlib, Stable Baselines, or OpenAI Gym; and foundational knowledge of mathematics and statistics, including linear algebra, probability and statistics, and optimization methods.

Preferred experience includes game AI development, practical projects involving automated game play or balance adjustment with reinforcement learning, distributed learning or development with AWS, GCP, Azure, or Kubernetes, and MLOps or DevOps work such as model deployment, continual learning, and CI/CD. Experience with imitation learning or data-driven AI using player behavior data, building simulation environments for agent training, and integrating AI tools into game-development workflows is also welcomed. English ability for reading technical documents and communication is listed as a preferred qualification.

The posting describes a preference for people interested in combining game development with AI, willing to learn new technologies, able to apply machine-learning or reinforcement-learning research to practical game development, and able to collaborate with game designers and engineers. It also mentions solving unfamiliar problems through trial and error and applying AI to improve player experience.

The employment arrangement is 業務委託. Compensation is listed as negotiable at JPY 250,000 to 900,000 per month. The work locations are offices in Shibuya and Meguro, Tokyo, and the location, workload, and compensation are to be discussed in an interview. Listed support programs include general meetings, an official club program, social-event expense support, a game-play room, onboarding training, CyStudy, Skill Thursday, influenza vaccination, in-house osteopathic treatment, free soft drinks, and a lunch subsidy.