You know that moment in a game when the AI does something and you think — wait, was that actually intelligent? That's what our team works on. We make games where NPC behaviour feels alive: enemies that adapt tactics based on observed player patterns, squad companions that adjust their role dynamically, and difficulty systems that stay in the flow zone without ever feeling like they're cheating or babying you. The underlying ML is real. We train models on player behaviour data, evaluate them with rigorous offline metrics, and stress-test them with human playtesters who are specifically trying to break them. It's applied machine learning where the feedback loop is short, the iteration is fast, and the results are immediately observable. We're looking for a junior ML engineer who has built something with machine learning — a game AI, a classifier, a recommender, anything — and can talk clearly about what worked and what didn't. The enthusiasm for games helps. The engineering discipline matters more.
Responsibilities
Build and evaluate ML models for NPC behaviour adaptation and player-facing difficulty systems with senior guidance
Process and analyse player behaviour data to identify patterns and features for model training
Write clean, tested Python code for model integration with the game engine
Participate in weekly playtesting sessions and document AI behaviour observations clearly
Collaborate with senior ML engineers on experimental design, model evaluation, and iteration
Requirements
Python — clean and logical; your functions do one thing, your variable names mean something, and you'd be comfortable showing your code in a review
Foundational ML knowledge: training, evaluation, overfitting, and the difference between model capacity and data quality problems
NumPy and Pandas for data processing and feature engineering
Any hands-on experience with Scikit-learn or PyTorch — at least enough to build and evaluate a model yourself
SQL for pulling and exploring data from our game analytics warehouse
Genuine interest in games — you play them, you notice how the AI feels, and you have opinions about it
Benefits
Work on game AI problems that are immediately visible, playable, and improvable
Full remote with optional studio visits (London)
$62,000 – $82,000 base salary
Game library budget: $300 per quarter — yes, this is legitimate research
Small team, flat structure, fast iteration — your experiments show up in a build within days