At Target, I invented and patented CORE, a contextual-bandit offer recommender, and took it from research to production. It drove a 70% increase in redemption and a 28% increase in opt-in, worth $500M+ in incremental sales.
I also shipped a real-time sponsored-product relevancy classifier that lifted relevancy by 3.8% in an A/B test while holding p95 latency to 10ms on CPU at 10k QPS. Ranking teams adopted its latency budget as a benchmark.
For Target's search and forecasting work, I designed an LLM labeling pipeline for more than 10 million query-product pairs and built an audit framework adopted by other workstreams. I led TargetRun, which cut MAPE/MAE by 50%, and architected Lookalike, a RAG system that reached 100% coverage at K=3 and removed eight manual hours per campaign.
Earlier, at TATA AIA Life Insurance, I applied NLP to onboarding and built a profit-prediction model; at HDFC Life Insurance, I built production systems for face recognition and lead allocation. I work across research and production, including ML platforms and evaluation frameworks used by teams beyond my own.

