AI Engineer Intern

• Built a TurboVec index retrieval + Two-Tower recommendation system serving 300K+ users with less than 100ms latency
• Architected a Two-Tower ranking pipeline using 100+ behavioral and graph features, improving Recall@100 and NDCG@10 by ~18%
• Implemented evaluation frameworks using Recall@K, NDCG, and MRR to benchmark retrieval and ranking models
• Automated evaluation and benchmarking across 10+ ranking and grading metrics, accelerating experimentation by ~3×
• Engineered CLIP-based embeddings for places, restaurants, and experiences, supporting retrieval across 500K+ items