About Us
We are a VC-backed startup focused on hyper-personalisation, currently in stealth. Inspired by the latest in recommender systems, we leverage transformers and graph learning alongside decision-making models to build the most engaging customer experiences for in-store retail.
Our mission is to change retail forever through hyper-personalised experiences that are both simple and beautiful.
About the Role - Offline Contextual Bandits and RL for Hyper-personalisation
We are looking for a Research Engineer to build decision-making models for in-store hyper-personalisation, with an initial focus on learning from logged human interaction data in an offline setting. You will work closely with domain experts and engineers to develop contextual bandit and reinforcement learning approaches that can support both single-step decisions and multi-step customer journeys, with the potential to enable online learning over time.
Key Responsibilities
Develop and productionise offline contextual bandit and offline RL methods that learn from logged interaction data.
Build rigorous off-policy evaluation (OPE) and counterfactual validation to measure candidate policies offline and compare approaches reliably.
Formulate and model both single-step decisions (contextual bandits) and multi-step decision processes (sequential / RL style settings) based on real retail interactions.
Advance representation learning for decision-making, including using transformers and GNNs where appropriate for behavioural, relational, and sequential data.
Translate research ideas into robust systems: dataset design, modelling, evaluation, deployment, monitoring, and iteration.
Collaborate cross-functionally to turn ambiguous product goals into concrete ML objectives, experiments, and deliverables.
Essential Qualifications
3 to 5+ years applying machine learning research in production settings.
MSc in Computer Science, Machine Learning, or a closely related field (or equivalent experience).
Strong foundations in machine learning and deep learning, including experience with at least one of: contextual bandits, reinforcement learning, counterfactual learning, ranking, or recommender systems.
Excellent Python skills and experience developing and debugging production-level code.
Ability to reason about evaluation methodology and failure modes when learning from logged interaction data.
Desired Skills (Bonus Points)
Demonstrated experience with offline policy learning and evaluation methods (for example IPS style estimators and doubly-robust approaches, plus uncertainty estimation).
Familiarity with bandit algorithms and exploration strategies, with interest in enabling online learning when the product is ready.
Experience with recommenders and ranking (candidate generation, reranking, slates).
Experience building data pipelines and improving data quality in modern ML environments.
PhD in a relevant field.
What We Offer
Opportunity to build technology that will transform millions of shopping experiences.
Real ownership and impact in shaping product and company direction.
A dynamic, collaborative work environment with cutting-edge ML challenges.
Competitive compensation and equity in a rapidly growing company.
If you’re excited by the idea of shaping the future of retail and eager to make a real impact from day one, we’d love to hear from you.
algo1