I work on Rust, SQL, and machine learning at the University of Fribourg with the Earth Metabolome Initiative, am affiliated with Lawrence Berkeley National Laboratory, and am an Apache DataFusion committer. I am one of the main contributors to Diesel, sqlparser-rs, and bhtsne.
On the GPU side, I contribute to the Burn ecosystem, including quantized matrix multiplication, FFTs, and atomics. While working on compiler correctness, I isolated a ptxas miscompilation that silently produces wrong results at any non-zero optimization.
Within the Earth Metabolome Initiative, I build Rust libraries for cheminformatics and mass spectrometry, among them smiles-rs, molecular-formulas, and mascot-rs. I have contributed spectral similarity algorithms to matchms. I also develop geometric-traits, a library for graph algorithms and combinatorial optimization such as maximum matching, minimum cost assignment, planarity testing, and subgraph isomorphism.
During my PhD I was the main contributor to GRAPE, ensmallen, and embiggen, a Rust and Python ecosystem for large scale graph processing and representation learning.
I am building connetto-rs, a local-first Rust stack where native and browser applications work against local SQLite replicas while PostgreSQL remains authoritative. It provides offline operation, live query subscriptions, and authorization-aware change delivery without a network round trip per interaction, across server, native client, and WASM client crates. I am also the main contributor to OpenScientist, an autonomous hypothesis testing agent for scientific data, where I built the agent layer, the model provider integrations, and the air-gapped execution mode.