Insider One - Senior Machine Learning Engineer (AI Decisioning)

• Design, build and release real-time decisioning systems that learn from interaction feedback at scale. • Own the modeling logic from how we represent users and signals to how reward attribution closes the loop and improves the next interaction. • Move ideas from papers and notebook prototypes including online learning policies and counterfactual estimators to production code that runs behind real-time APIs. • Design and build resilient streaming/batch pipelines that feed user state, reward signals, and offline replay. • Run honest experiments, A/B test what you ship, design offline evaluation for what you cannot, and kill your own work. when it doesn’t outperform against the baseline. • Continuously monitor and improve the quality, latency, observability, and scalability of the systems. • Collaborate across platform and product teams to turn research-grade ideas into production-grade products. • You don’t need to tick every box • if you likely have: • Designed and deployed personalization, ranking, or recommendation systems used by real users; improved core engagement metrics (CTR, conversion, retention, revenue) and can talk concretely about the lift • Familiarity with sequential recommendation, ranking, or joint / multi-objective optimization problems where you can't optimize one metric without trading off another • A/B tested the impact you’ve provided and iterated based on what the data said • Comfortable with the messy reality of production ML: cold start, sparse signal, label delay, feedback loops, distribution shift • Solid grounding in probabilistic modelling (Bayesian inference, calibration, hierarchical models) and modern recommender techniques (embeddings, sequence models, LLM-driven content understanding), applied to sequential, ranking, or multi-objective problems • Software engineering, production-quality code and at least one programming language, care about API contracts, testing, and observability, not just notebooks • Built high-throughput real-time or batch pipelines supporting ML training and inference, on AWS (or an equivalent major cloud) comfortable owning a service end to end across compute, storage, networking, and CI/CD • Have moved at least one model from a paper, a notebook, or a whiteboard sketch into a real system that serves traffic, and can speak honestly about what broke along the way • and it would be a strong plus if you have any of below: • Hands-on experience with online decision-making under uncertainty, multi-armed bandits, contextual bandits, Thompson sampling, UCB, or RL agents that have served traffic • Comfortable reasoning about exploration vs exploitation, regret, off-policy evaluation (IPS, doubly robust), counterfactual estimation, and the failure modes of each • Causal inference / uplift modeling • Academic research experience or publications in online decision-making under uncertainty, reinforcement learning, sequential recommendation, optimization, or related fields.