High-Throughput, Cost-Effective Billion-Scale Vector Search with a Single GPU [to appear in SIGMOD'26]
-
Updated
Jan 16, 2026 - Cuda
High-Throughput, Cost-Effective Billion-Scale Vector Search with a Single GPU [to appear in SIGMOD'26]
GPU-native approximate nearest neighbor search for Apple Silicon. Pure Swift + Metal — CAGRA-style graph index with full mutability, filtered search, streaming ingest, and multiple persistence modes.
Temporal and category-optimised graph ANN index supporting “top‑k neighbours of q with category C in time window [t_start, t_end]” via category-aware Filtered‑Vamana and Historic Neighbour Tree.
A multilingual FAQ system using Sentence-Transformers + Pinecone semantic retrieval outperforming keyword-based baselines on low-latency, high-concurrency benchmarks.
Add a description, image, and links to the annsearch topic page so that developers can more easily learn about it.
To associate your repository with the annsearch topic, visit your repo's landing page and select "manage topics."