In this AI Tech Experts Webinar, Grzegorz Rybak (Senior Data Engineer) and Cezary Gorczyński (Data Engineer) explore whether #DuckDB + #DuckLake can serve as a practical lightweight data platform for modern analytics workloads.
They walk through common consulting scenarios where teams must balance fast delivery with long-term scalability and explain how a DuckDB-based stack can provide strong analytical performance without committing to a full warehouse platform. Topics covered
real trade-offs:
- concurrency limits, streaming gaps, file fragmentation
- DuckDB as an embedded OLAP analytical engine
- querying Parquet directly with a zero-ingest workflow
- DuckLake as a lakehouse management layer
- the “Holy Trinity” architecture: compute, storage, metadata
Timeline
00:00 DuckDB intro
00:57 Client problems: greenfield vs legacy data stacks
06:42 DuckDB architecture and in-process analytics engine
12:38 DuckLake lakehouse layer and the “Holy Trinity” architecture
17:23 Trade-offs: concurrency, scaling and streaming limits
21:37 Conclusions: portable lakehouse strategy
Speaker
Grzegorz Rybak
Senior Data Engineer at deepsense.ai
Cezary Gorczyński
Data Engineer at deepsense.ai
