Industry Collaboration Project - SAMSUNG
ARSHAD NADEEM KHAN
Industry Collaboration Project - SAMSUNG PRISM
Development Of Multimodal Vector Database Using Embedding Vectors
• Developed a multimodal on-device vector search system to enable semantic retrieval across image, text, and audio data.
• Implemented CLIP-based embedding pipeline with FP16 compression and optimized HNSW indexing (hnswlib) for fast nearest-neighbor search on mobile devices.
• Achieved -95% Recall@10 with sub-millisecond search latency, enabling efficient and scalable on- device semantic search for large datasets.
Like this project
Posted Apr 17, 2026
Industry Collaboration Project - SAMSUNG PRISM Development Of Multimodal Vector Database Using Embedding Vectors • Developed a multimodal on-device vector se...
Likes
0
Views
0
