Lead Data Engineer · 9+ years

Aleksandr Andreev

I build streaming, lakehouse and applied-AI platforms at production scale — Kafka, Flink, Spark, Iceberg, Trino and the tooling that keeps them reliable.

Selected work

Project

GitLab MR Review Agent

LLM-powered code review bot using FastAPI, Qdrant, Claude API, and RAG over Confluence. Reduced review turnaround by 40%.

llm rag fastapi python qdrant

3 min read

Article

DuckDB httpfs column pushdown over Parquet on S3

DuckDB pushes column projection and filter predicates down to S3 Parquet reads — but only if your file stats are valid.

duckdb parquet s3

1 min read