The Role:
The game-generation pipeline is the product. Every game in the feed comes out of it: a creator's prompt goes through intent extraction, template matching, model routing, asset retrieval, generation, validation and repair, and finally the runtime that plays it.
You will lead the team that builds and runs this system and own its design. You set the architecture, the roadmap, and the evaluation strategy, and decide where models, templates, and code each do the work. The team is measured on two numbers: playability pass rate going up and cost per generation going down, at volume.
This is a leadership and system-design role. Your time goes to architecture, technical direction, prioritization, reviews, and hiring. You should be able to prototype an idea or read through a failing generation trace yourself, but you will not be writing production code day to day.
Responsibilities:
- Lead the AI team (ML/data engineering, game development, and a contracted engineering partner): set priorities, run planning, unblock people, and grow the team as we scale.
- Own the architecture of the generation pipeline end to end and decide how it evolves.
- Define how we measure quality: benchmark suites, LLM-as-judge, and playability pass/fail checks, and use them to show output improving over time.
- Design the feedback loop that brings play data (completion, replay, shares, remixes) back into generation.
- Own the economics: model selection and tiered routing, templates and retrieval, caching, and when distillation or fine-tuning is worth it.
- Work with product, game design, and core engineering to turn what creators and players need into a technical roadmap.
- Keep us current on models and tooling, and make the call on what to adopt.
Requirements:
- Have led an engineering team that shipped an LLM-powered generation system to production at real volume.
- Have designed multi-stage generation systems with evaluation built in: generation, validation, repair and eval, and the feedback loop between them.
- Understand the cost side well: you have made model routing, caching, and template/retrieval decisions that materially cut cost per generation while holding quality.
- Technical enough to hold your own in architecture and code reviews, and to prototype when it is the fastest way to settle a question.
- Comfortable making product trade-offs as well as technical ones.
Preferred Qualifications:
- Experience in games. You have shipped games, worked at a games studio, or built tooling or content systems for games, and you know what makes a short game fun to play.
- Experience with code generation or prompt-to-app systems.
- Hands-on work with game engines or web game frameworks (Unity, Godot, Phaser, Three.js), or procedural content generation.
- Fine-tuning or distilling open models in production.
- Image or asset generation experience.
- Startup or founding-team experience, or having built a team from a small base.
