AI Engineer Intern

• Architected an end-to-end personalized discovery feed (retrieve → rank → re-rank) for a 188K-item catalog using hybrid Elasticsearch BM25 and vector retrieval with slot-based re-ranking
• Designed a recency-decayed user preference model to personalize retrieval and support future Two-Tower and SASRec ranking
• Reduced cold-feed latency from 8s to <500 ms via three-tier caching, cache pre-warming, optimized indexing, and geospatial retrieval
• Built recommendation analytics using RudderStack to capture user behavior and optimize personalization through engagement metrics
• Resolved ES-MongoDB data consistency issues and hardened location-aware retrieval, improving feed and recommendation reliability
• Engineered the recommendation architecture to support future Two-Tower and SASRec models