Machine Learning Engineer, Unity

• PhD in Computer Science, Machine Learning, Statistics, or a related field (graduating 2026 or recent graduate)
• Strong research foundations in one or more of: recommendation systems, reinforcement learning, LLM post-training or alignment, human-AI collaboration, probabilistic modeling, or optimization
• Experience working with large-scale data and ML systems, whether through research or industry internships
• Fluency in Python; familiarity with ML frameworks such as PyTorch or TensorFlow
• A track record of rigorous, high-quality research — publications at top venues (NeurIPS, ICML, ICLR, KDD, RecSys, ACL, WWW, or similar) are a strong signal
• Strong written and verbal communication skills — able to make complex ideas accessible across technical and non-technical audiences

Desirable

• Unity's Vector AI team builds the machine learning systems that decide which ads reach which players — across billions of monthly users on the world's leading game engine
• Recommendation and ranking systems are the core of this work: predicting user value, optimizing bids, and delivering outcomes for advertisers at massive scale
• We are building the next generation of these systems. The frontier has shifted — large language models, reinforcement learning from human feedback, and agentic AI are reshaping what recommendation systems can do
• We are looking for PhD graduates who have worked at that frontier and want to bring those ideas into production systems that matter
• Design, build, and evaluate next-generation ranking and recommendation models that incorporate LLMs, RLHF, and preference learning to improve ad relevance and user experience
• Develop user understanding systems — conversion prediction, behavioral modeling, and value estimation — that operate across billions of impressions
• Apply reinforcement learning and optimization techniques to bidding strategy, auction dynamics, and real-time ad delivery
• Design and run rigorous experiments using causal inference, A/B testing, and offline evaluation frameworks to measure and improve model quality
• Partner with engineering to bring research ideas into production, working across the full pipeline from training data to deployed model
• Communicate findings clearly to technical and non-technical stakeholders across engineering, product, and business teams