Senior Recommendation / Growth ML Engineer

• Design and build low-latency real-time recommendation systems that personalize trading product discovery, content feeds, and community content ranking (ByX) for users across web and mobile surfaces — covering the full ML lifecycle from data preparation and feature engineering to model training, evaluation, and production deployment; campaign targeting logic, subsidy decisions, and customer-facing launch approvals are owned by offshore growth teams.
• Apply advanced ML personalization techniques — including two-tower retrieval, sequential models, graph-based methods (GNN), multi-objective modeling (PLE/MMoE), and contextual bandits — to deliver highly relevant and engaging experiences across Bybit's trading and social surfaces; explore multi-scenario joint modeling to unify signals across trading, community, and campaign surfaces
• Build the ML infrastructure and model research layer for AI-powered personalization for real-time ranking and retrieval systems.
• Build the recommendation and experimentation infrastructure for user lifecycle management; US persons are excluded from any targeting universe.
• Develop predictive models for user churn, upgrade propensity, reactivation likelihood, and LTV — applying causal inference (uplift modeling, difference-in-differences) and operations research methods.
• Build and maintain real-time and batch feature pipelines that feed recommendation and growth models; partner with data engineering on feature store design; ensure end-to-end system observability and debugging tooling for production recommendation services
• Partner closely with Growth Product, Data Science, ByX Community, and Asia-Pacific engineering teams to define success metrics, translate business goals into ML system requirements, and ship measurable impact; define engineering standards, conduct design reviews, and mentor junior engineers as the US team grows.

Major Requirements

• 5+ years of industry experience in ML engineering, recommendation systems, or growth engineering at a consumer-scale internet company
• Proven track record building and shipping real-time recommendation or personalization systems serving millions of users; strong knowledge of recommendation algorithms including collaborative filtering, two-tower models, sequential models, graph-based methods (GNN), multi-objective modeling (PLE/MMoE), and reinforcement learning / contextual bandits
• Build high-throughput real-time feature pipelines (Kafka/Flink) enabling minute-level user behavioral feature updates; contribute to a unified online/offline Feature Store architecture, governing feature consistency and eliminating time-travel leakage across training and serving.
• Own the construction and optimization of large-scale vector retrieval systems (Faiss/Milvus/HNSW) supporting candidate pools scaling from thousands to millions of heterogeneous items (trading products, news, KOL content, on-chain signals).
• Strong proficiency in Python and at least one JVM or compiled language (Java, Scala, Go, C++); experience with ML frameworks (PyTorch, TensorFlow, or JAX); proficiency in big data tools (Hive SQL, Spark, Flink, or MapReduce)
• Ability to collaborate effectively with Asia-Pacific engineering and product teams in Mandarin Chinese.
• Nice-to-have:
• Hands-on experience with large-scale data infrastructure: Kafka, Spark/Flink, Redis, feature stores, and online serving systems.
• Experience in crypto/Web3 or fintech with strong understanding of user behavior in financial contexts; experience with LLM-based personalization or generative recommendation architectures; experience in multi-scenario joint modeling (unifying signals across search, recommendation, and marketing); experience with LTV prediction, operations research, or subsidy/budget optimization; experience building recommendation systems for social/community platforms; publications at top AI/ML venues (KDD, NeurIPS, WWW, SIGIR, WSDM, CIKM, ICLR, ICML).