Abylay Amanbayev

Teaching/Research Assistant

University of California Merced

Public research university in Merced, California.

Researches query and retrieval optimization, benchmarks vector databases, built a multimodal benchmark dataset, and teaches database and computer architecture courses.

• Current work: Designing a query optimizer for Filtered Approximate Nearest Neighbor Search and a retrieval optimizer for agentic retrieval — selecting which backend to query (vector, relational, lexical, etc.). • Filtered Vector Search: Engineered and benchmarked Vector Databases (PG-vector, Milvus, and FAISS) across vector search indexes (including HNSW, IVF); developed the new metric to predict query latency.

Founder & Sole Engineer

dalel.ai

Built a cited-answer chatbot for Kazakhstan tax legislation and engineered its retrieval tools, backend, and frontend solo.

• Built a chatbot over the Kazakhstan tax legislation: indexed the full large text corpus for exact document lookup plus Chroma vector search for semantic retrieval, returning cited answers. Demo: letmesee / dalelai • Designed the kaztax MCP server (FastMCP): a model-agnostic tool layer the agentic LLM (Gemini / Claude / GPT) drives to invoke those fetch and search methods, returning cited, grounded answers. • Engineered the full stack solo — PostgreSQL with async SQLAlchemy + Alembic, FastAPI with structlog and rate limiting, Next.js / React frontend, deployed via Railway.

Team Lead, Data Science Challenge

Lawrence Livermore National Laboratory

U.S. national laboratory focused on science and technology for national security.

Led a four-person team applying deep learning to biomedical signal reconstruction.

• Led a 4-person team applying deep learning to biomedical signal reconstruction.

PhD, Computer Science

University of California Merced

Master, Data Science

HSE University