Senior Game AI Developer – Custom Reinforcement Learning Pipeline
Company: Leading Organization
Location / Work Mode: Remote / Global
Employment Type: Full-Time
Salary / Compensation: Competitive / Negotiable
Apply NowRole Overview
We are seeking an exceptionally talented and experienced Software Engineer with a strong background in machine learning and game engine development. This is a unique opportunity to design and implement a cutting-edge, custom self-learning AI agent for the acclaimed video game, Overgrowth (available on Steam). If you're passionate about pushing the boundaries of game AI and working with open-source engines, this role is for you.
Leveraging Wolfire Games' decision to make Overgrowth's core engine fully open-source, this project aims to revolutionize NPC behavior. Instead of relying on traditional hard-coded scripts, you will implement a true Reinforcement Learning (RL) training framework. Your mission will be to train an autonomous NPC agent that dynamically adapts, learns from combat outcomes, and continuously optimizes its strategy to evolve into an elite, formidable fighter within the game world.
This is a chance to make a significant impact on game AI, contributing to a project that blends advanced machine learning techniques with real-time game development. You will be at the forefront of creating intelligent, adaptive game characters that offer an unparalleled player experience.
Key Responsibilities & Duties
- Engine Integration: Clone and meticulously configure the open-source Overgrowth C++ repository, then modify the NPC character controllers and physics loops to expose critical real-time combat telemetry, including spatial coordinates, bounding boxes, health pools, velocity, and attack frames.
- Networking Pipeline Development: Build a robust localhost communication bridge, utilizing technologies like UDP/TCP sockets or shared memory, to efficiently stream real-time telemetry data out of the game engine into an external environment.
- ML Framework Development: Create a companion Python environment, leveraging PyTorch and Gymnasium, to consume the incoming data stream, manage complex reward matrices (scoring hits, blocks, positioning, and avoiding damage), and send precise discrete control inputs back into the game engine to dictate agent actions frame-by-frame.
- Compilation & Deployment: Successfully configure and compile the final customized game build using CMake and Visual Studio (MSVC) on the Windows platform, ensuring stability and performance.
- Handoff & Optimization: Provide comprehensive, clean documentation for the developed systems and assist with the initial setup and training deployment via a remote session to ensure seamless operation on the client's hardware.
Requirements & Qualifications
- Strong proficiency in C++ programming and practical familiarity with game physics and actor loops.
- Demonstrated experience with CMake build systems and compiling open-source game engines.
- Proven background in Machine Learning, specifically with Reinforcement Learning (e.g., PPO, DQN, or Actor-Critic architectures).
- Expertise in Python, PyTorch, and the Gymnasium/Open AI Gym ecosystem.
- Experience with network socket programming (UDP/TCP).
Benefits, Perks & Culture
- Competitive Compensation: Enjoy a competitive and negotiable salary package that reflects your expertise and contributions.
- Global Remote Flexibility: Work from anywhere in the world with a flexible remote setup that supports work-life balance.
- Cutting-Edge Technology: Engage with advanced Reinforcement Learning and game engine development, pushing the boundaries of AI in gaming.
- Impactful Work: Contribute to an exciting open-source project and directly influence the future of game AI.
- Professional Growth: Opportunities for continuous learning and development in a highly specialized and innovative field.
How to Apply
Ready to take on this exciting challenge? If you meet the qualifications and are eager to make a significant impact on game AI, we encourage you to apply directly through our application portal.