AI/ML Engineer — Level Generation & Solver

About the role

Every one of OutBrick’s 2,000 boards was cleared by a solver before it shipped, so no player ever meets a dead end. You will own that pipeline and push it further: generating boards that are interesting as well as solvable, grading their difficulty, and setting move targets that feel fair.

You will also explore personalisation — pacing difficulty to the player in front of the board — under one firm constraint: OutBrick has no analytics SDK and does not collect gameplay data, so anything that adapts to a player has to run on the device.

What you’ll do

• Own and extend the board solver and its verification of every shipped board
• Build procedural generation for new chapters, with difficulty grading and move-target setting
• Research on-device personalisation of pacing and hints that needs no data to leave the device
• Build tooling so designers can generate, inspect and tune boards themselves
• Keep the pipeline reproducible and the results explainable to a non-specialist

What you’ll bring

• Strong software engineering in Swift, Python, C++ or Rust — and willingness to work across them
• Experience with search and planning (A*, IDA*, BFS/DFS with pruning, constraint solving) or combinatorial optimisation
• Practical machine-learning experience, from data to evaluation
• The judgement to choose a simple heuristic over a model when the heuristic is better

Nice to have

• Experience with procedural content generation for games
• On-device ML on Apple platforms (Core ML, Create ML, Metal)
• A published paper, talk or open-source project on puzzles, search or PCG

Don’t tick every box? Apply anyway. We would rather meet someone who can grow into this than miss them.