The opportunity
Instacart’s Shopping Experience team is focused on making it fast and effortless for customers to find the right items within a single retailer and complete their order with confidence. As a Senior Data Scientist dedicated to Search, you’ll own the analytics and experimentation…
What you'll do
Own core Search metrics and funnels end to end (e.g., query → impression →: engagement → cart adds), including defining guardrails, monitoring performance across platforms and segments, and diagnosing conversion gaps.
Design, run, and interpret experiments across ranking, retrieval, and search: UX (e.g., relevance model changes, query understanding, result layouts), turning ambiguous or conflicting outcomes into crisp, data-driven recommendations.
Partner with Product, Engineering, and ML to prioritize opportunities, size: impact, and influence the roadmap for relevance, quality, and latency improvements that unlock measurable business outcomes.
Build deep diagnostic analyses by query class, price point, surface, and: customer lifecycle to pinpoint where and why Search underperforms and specify concrete changes that will move key outcomes.
Connect offline model evaluation with online and business metrics by: collaborating with ML partners on evaluation design, ensuring model changes reliably improve end-user experience—not just offline scores.
Improve data quality, instrumentation, and metric definitions for Search so: that teams can reason about performance with clarity, consistency, and speed.
What they're looking for
- Experience in search relevance, ranking, recommendations, personalization, or: information retrieval (e.g., e-commerce or marketplace search).
- Familiarity with NLP, embeddings, and semantic search, including how to: evaluate and iterate on these techniques in production.
- Experience bridging offline evaluation metrics (e.g., NDCG, precision/recall,: human evaluation) with online experiments and business outcomes.
- Background in causal inference beyond standard A/B tests (e.g., holdouts,: diff-in-diff, quasi-experiments) to measure long-term or cross-surface effects.
- Comfort working across web and native app surfaces, navigating tradeoffs: between relevance, monetization, and latency.
- Proven impact improving logging, instrumentation, and metric definitions in complex data environments.
