High-Performance ML & Computer Vision · Heterogeneous AI Systems
I'm a machine learning engineer focused on computer vision and inference at scale — taking models from research code to production across GPUs, edge accelerators, and everything in between. I care about the parts most people skip: latency budgets, memory footprints, and pipelines that keep working after the demo ends.
Outside of shipping, I write about deployment tradeoffs and heterogeneous compute on the blog below.
Free removes cost, not risk: Console, OpenRouter and NVIDIA free models compared on privacy, geoblocking, rate limits, and safe fallbacks.
Choose between one agent, pipelines, parallel workers, and bounded delegation using task dependencies, acceptance criteria, and coordination costs.
Prototype CUDA projects on Colab, then move to persistent cloud GPUs for larger workloads and reproducible development.
Happy to talk about computer vision and ML infrastructure.