Software Development Engineer - OpenSearch, VectorSearch, OpenSearch

Software Development Engineer (SDE II) - Vector Search Team

The Open Search Vector Search team builds and operates the core vector search engine that powers recommendations, semantic search, and multi-modal search capabilities across Amazon's flagship services including Amazon Prime, Amazon Retail, Amazon Music, and Alexa. Our technology enables Amazon Open Search Service to handle large-scale vector search operations at billions of high-dimensional vectors, contributing to the advancement of vector databases and AI/ML infrastructure.

We are at the forefront of building the retrieval and reasoning substrate for modern enterprise AI systems.

We are looking for a passionate and talented Software Development Engineer (SDE II) to join our Vector Search team. You will design, build, and optimize high-performance vector search systems that operate at massive scale. This role offers the opportunity to work on cutting-edge problems in approximate nearest neighbor (ANN) search, GPU-accelerated indexing, new algorithms to improve the recall and performance — all within one of the most widely adopted open-source search platforms in the world.

Key job responsibilities:

• Design and implement high-performance vector search algorithms (HNSW, cluster-based ANN, hybrid approaches) for billion-scale datasets
• Develop and optimize GPU-accelerated vector indexing and search capabilities
• Build and enhance the k-NN plugin for Open Search, including support for new distance metrics, quantization techniques, and memory-efficient graph structures
• Contribute to tiered storage architecture for vector data across hot/warm/cold tiers to improve cost efficiency
• Implement memory-optimized approaches such as MMAP-based graph loading and partial graph loading to reduce working set size
• Collaborate with applied scientists on novel algorithmic approaches to vector search optimization
• Participate in open-source community contributions, code reviews, and technical design discussions