Engineering
Systems-level engineering, selected work, and reproducible claims.
How we work
Performance claims are reported as worst-of-10, not best-of-10. Five warmups, twenty timed iterations per run, ten runs. Numbers are reproducible on the documented hardware.
These are not optimized benchmarks. They are reproducible baselines.
Property-based testing where it matters. CPU reference checks for every kernel. Independent security validation for every shipped cryptographic system.
Engineering breadth
Over 15 years of production engineering across systems languages, low-level performance work, and full-stack platforms. The work predates the current AI tooling cycle.
lines authored
commits
languages, deep production
Python
Rust
Rust + Python footprint
Tokei against on-disk repositories, consolidated language buckets. On-disk only — does not include archived or remote-only work.
Production deployments at
Open work
Kaio — GPU kernel authoring in Rust
A framework for writing and compiling GPU kernels directly to PTX, designed for control over memory, execution, and performance.
- Rust-native
- PTX compilation at build time
- memory control
- no CUDA toolkit dependency