Ernest

This is a publication, not a portfolio. The work is compilers, codecs, inference, and the measurements that decide whether a change was real.

I build systems and then write down what they actually did. A kernel that is faster in isolation is not a result until the pipeline has paid for it. A fallback that produces an answer is not a refusal. A jitter buffer built for the public internet is not a latency budget for a quiet LAN.

The essays are the long form. The lab is the notebook: dated, shorter, and allowed to record a failure without turning it into a lesson with a bow on it. Projects are the artifacts those notes come from. Releases are dated, and only the ones that changed something get a page of their own.

I am not trying to be a company, a course, or a feed of takes. If a number cannot name its machine, its baseline, and what it failed to survive, it does not get published here.