03/2026 BIFOLD Colloquium

CedarDB: Bringing a code-generating research system into production

Abstract:

CedarDB is engineered from the ground up for modern hardware, combining transactional and analytical workloads without compromises. At its core is query code generation: instead of interpreting SQL at runtime, CedarDB compiles each query into optimized machine code, fusing operators into tight pipelines and parallelizing across all cores using morsel-driven parallelism. But building a fast research system is only half the story. This talk draws on the real-world experience of transforming Umbra, a TU Munich research project, into CedarDB. It shifts priorities from benchmark-driven development to tripling the test suite, handling unexpected system setups, and building the observability needed to understand how users actually run your system. Combining the solid technical foundation of Umbra with the reliability-driven engineering at CedarDB is the key to our development.

Short-Bio

Maximilian Bandle is a Founding Engineer at CedarDB, where he brings together technical expertise in database systems and a hands-on entrepreneurial background. He received his Ph.D. from the TU Munich, where his research focussed on code-generating database systems and data processing on modern hardware. He published papers about join processing, filtering, and query compilation at VLDB and SIGMOD.
Alongside his academic career, Max taught courses on turning prototypes into real-world products and co-founded Mintech, a prototyping-as-a-service company helping clients bring their ideas to life across hardware and software. He has since returned to the technical frontier, contributing to CedarDB's growth as both a product and a company.