Different roles
I’ve enjoyed working in.

I study when AI-generated systems code deserves to be trusted. I built the agent methodology, compiler and linker analysis tooling, and rebuilt a legacy fuzzer as a concurrent, resumable, checkpointed Go tool used to evaluate 10 Linux utilities against their human-written counterparts.

I wanted a stronger feel for the machinery beneath reliable services, so I joined a selective production-engineering fellowship and learned directly from Meta engineers. I applied those lessons by keeping a Flask/MySQL service stable on a memory-constrained VPS, shipping a second Go/SQLite service on its own host, and building the deployment gates and dashboards that tell you when reality disagrees with your plan.

I contribute to large open-source codebases and get changes through maintainer review: Google's Gemini CLI, NVIDIA's Sirius DB, GNU coreutils, less, and dash.

I learned to start with the environment, not the model. In an understaffed maternity hospital, that meant code-switched speech, noisy rooms, and clinicians who needed less paperwork, not another tool to manage. The result ran in production for seven months.

Teaching at scale turned me into an infrastructure engineer. I led a 12-person team and rebuilt the bottlenecks that surfaced on deadline days: an autoscaling AI tutor, a 12× faster grader, and observability that made failures diagnosable before they became course-wide problems.

I ship my own tools end to end (design, implementation, tests, CI, releases, and the support that follows) so the work is judged by how it holds up, not by how the demo looked.