constructor - Senior Product Manager: Recall

• Set a unified roadmap: Define the multi-quarter vision and strategy across both the Machine Learning & Recall and Query teams, ensuring query understanding and candidate generation evolve hand-in-hand. • Drive Query intelligence: Lead the Query team to push boundaries in tokenization, spell correction, entity extraction, intent classification, and LLM-assisted query parsing across multiple languages. • Advance Recall systems: Lead the Machine Learning & Recall team in evolving candidate generation architecture, seamlessly combining traditional keyword search with modern dense embeddings, vector retrieval, and hybrid recall models. • Move core business metrics: Own measurable lift in search-attributed conversion, revenue per visit, and GMV across all retail customer verticals. • Turn research into production wins: Work shoulder-to-shoulder with ML researchers, data scientists, and software engineers to take SOTA models from research to low-latency, cost-effective production systems. • Balance ML performance trade-offs: Make data-backed decisions balancing model complexity, accuracy, inference latency, compute costs, and real-time execution constraints. • Build evaluation frameworks: Establish robust offline and online measurement systems to evaluate retrieval precision and query understanding accuracy against customer revenue outcomes. • Orchestrate pipeline handoffs: Ensure candidate product sets and query context flow cleanly into downstream ranking and search quality models without signal loss. • Solve systemic issues: Lead technical triage when query interpretation or retrieval anomalies occur, building durable, automated fixes rather than one-off patches. • WHAT SUCCESS LOOKS LIKE • Evolved our hybrid recall framework to measurably improve candidate generation precision for long-tail queries and complex catalog structures. • Driven quantifiable lift in search-attributed conversion and revenue per visit across major customer verticals. • Stood up an integrated evaluation framework connecting query parsing and candidate recall directly to business outcomes. • Optimized inference latency and compute infrastructure to ensure high-capacity ML models run fast and cost-effectively in production.