Thomas Veasey

Principal engineer at Elastic, focused on vector search performance and quantization. Authored detailed benchmarks comparing Elasticsearch’s BBQ (Better Binary Quantization) with TurboQuant approaches.

Known For

• BBQ (Better Binary Quantization) implementation in Elasticsearch
• OSQ (Optimized Scalar Quantization) — 10-40x faster than traditional methods via integer SIMD
• Benchmarking quantization approaches on CPU hardware
• Anisotropic loss and centroid centering techniques

Articles

• Elasticsearch BBQ Optimized Scalar Quantization vs TurboQuant
• Elastic Learned Sparse Encoder (ELSER) Retrieval Performance

Key Concepts

• Vector Quantization
• Dense Vector Retrieval
• Approximate Nearest Neighbor Search

Affiliations

• Elastic