Software & mathematics
I'm a software engineer specializing in distributed systems, database internals, and correctness.
Previously at ScyllaDB and poolside, I've worked on consensus protocols, database control planes, and infrastructure for training programming models. I'm also interested in programming languages and the connections between mathematics and software, including AI-assisted proof-complexity research.
An open research notebook investigating superpolynomial proof-size lower bounds in restricted proof systems. The work is AI-assisted, and the central lower-bound question remains open.
The software supports reproducible experiments: verification programs, resource-controlled execution, timing and evidence records, and automated notebook publication through GitHub Actions.
A distributed key-value store with strictly serializable transactions, built on LPaxos: my replication algorithm for building fault-tolerant, linearizable services without replicated logs.
I developed the algorithm and implementation as part of my master's work in computer science. LattiStore includes a small transaction language and a runnable multi-node setup. It is a proof of concept; persistent storage is not implemented.
I contributed to Motions as part of my bachelor’s project: a Markov Chain Monte Carlo simulation of chromatin movements within a cell nucleus, modelling DNA–protein interactions. The project brings together mathematical modelling, scientific computing, and functional programming.
A Linux driver for the HardDoom II PCI device, implemented as a university operating-systems project. It handles asynchronous command submission, DMA buffers, interrupts, and device resource management.
An experiment in translating Haskell to C++ using Conal Elliott's “Compiling to Categories” approach. It explores the connection between functional programming, category theory, and code generation.
My thesis develops LPaxos, the distributed algorithm behind LattiStore. I wrote the thesis and developed the accompanying software; my advisor, Janina Mincer-Daszkiewicz, reviewed the thesis.
Co-authored with Krzysztof Pszeniczny, this thesis introduces persistent homology for computer scientists and data analysts: a way to study the shape of data by tracking topological features across scales.
Assuming only basic linear algebra, it develops the necessary topology and algebra, then explains persistence modules, barcodes, and an algorithm for computing persistent homology of a filtered simplicial complex.
I led a five-person engineering team responsible for much of the core database’s control plane, particularly schema and topology management. As a core maintainer with merge authority, I reviewed and gated contributions across multiple subsystems.
I worked on the containerization system used to train and evaluate foundation models for programming with reinforcement learning.
At ScyllaDB, I worked on the core of a distributed database: moving cluster management to Raft consensus, improving query performance, and supporting topology changes in Change Data Capture. I also built testing tools to find correctness bugs in distributed algorithms and used TLA+ to specify parts of the control plane.
I hold an M.Sc. in Computer Science (2020), a B.Sc. in Mathematics (2018), and a B.Sc. in Computer Science (2016), all from the University of Warsaw. I also studied mathematics at the University of Copenhagen on an Erasmus exchange in 2017–2018.
Based in Warsaw, Poland.
kamilgbraun@gmail.com LinkedIn Personal GitHub ScyllaDB GitHub CV (PDF)