We are building a high-load DBMS where full-text search, vector search, and analytics coexist within a single system. The analytical engine and storage are built on DuckDB, and search is powered by IResearch, our own C++ engine that the core team has been developing for years.
We are looking for a strong C++ engineer for the core team to work on the storage layer, query execution in DuckDB, and the search engine. This is not about "adding a feature on the side" — it's about diving into the very core.
Tech Stack: modern C++, Linux, DuckDB, full-text and vector search, vectorized execution, columnar storage, inverted indexes, multithreading and concurrency, performance optimization, low-level profiling.
What you'll be doing:
- Deeply understand the codebase: storage, query engine, search layer. Reverse engineering architecture, identifying bottlenecks, documenting key components.
- Develop the DuckDB layer: query planning and execution, storage, catalog/store, upstream bumps and extensions.
- Develop our search engine: inverted index and columnstore, scoring (BM25, WAND/MaxScore), vector indexes (IVF, quantization), tokenizers, posting list compression.
- Develop protocols and connectors: foreign servers, formats and sources (Parquet, Iceberg, CSV, JSON on S3/HDFS), compatibility with ClickHouse/Elasticsearch.
- Profile CPU/memory/disk, optimize critical sections, run benchmarks (ClickBench, ann-benchmarks, FTS).
- Write unit, integration, and load tests, participate in SQL fuzzing and code reviews.
Who we are looking for:
- Proficient in modern C++.
- 3+ years of systems programming experience.
- Strong understanding of algorithms, data structures, multithreading, and concurrency.
- Experience with high-load systems.
- Ability to understand complex third-party code.
- Understanding of operating system principles: memory, disk, network.
Will be a plus:
- DuckDB, ClickHouse, Elasticsearch/Lucene.
- Experience in database development, search engines, or query engines.
- Full-text search / IR (inverted index, BM25).
- Vector search / ANN (IVF, HNSW, quantization).
- Columnar storage, vectorized execution, query planning.
Working conditions:
- Flexible start to the workday;
- Possibility of hybrid or remote work format;
- Strong engineering team with whom you can learn a lot;
- Real impact on product architecture, not just backlog tickets;
- Voluntary Medical Insurance with dental coverage;
- Sports compensation;
- Foreign language learning compensation;
- Corporate training.
