Engineering that holds up in production

Rust-first, memory-safe systems

AI infrastructure that crashes in production is expensive. I use Rust, WebAssembly, and careful systems design to remove whole categories of failure before they reach users.

• High-throughput inference services and model routers
• Zero-allocation hot paths and latency optimization
• WebAssembly modules for browser and edge deployments

NVIDIA and GPU computing

From CUDA kernels to DGX Spark memory planning, I help you get real performance out of GPU hardware instead of leaving it on the table.

• CUDA optimization and unified memory planning
• Distributed training and inference orchestration
• Quantization, batching, and throughput tuning

Production, not prototypes

I have designed and shipped end-to-end AI systems from early prototypes to fully managed production deployments. Every line of code is written with observability, security, and maintainability in mind.