Generalist Software Engineer specializing in distributed systems. Over nearly a decade of professional experience, I've built a track record of delivering reliable, maintainable, and time-tested software, backed by rigorous, deterministic testing. My pull requests across the companies listed on my LinkedIn profile (including a unicorn and Big Tech companies) reflect this engineering discipline. I also believe in leaving a little room for safe experimentation—a small percentage of the time across a high volume (thousands) of commits over the past decade, with experimentation typically front-loaded when exploring new technologies or problem domains. However, maintaining low revisions per PR remains important, as it has in my career in Big Tech (~1.3). Regardless of code and PRs, my strength is in problem solving, and this has been demonstrated in projects, interviews (e.g. Senior Software Engineer @ Google), and more.
Self-taught. No prior education. No training (since 2015, but very informal). Little to no questions on the job. Look into it.
High amount of quality * quantity (not just total commits or lines of code) over ~10 years, thanks to my great work paradigms. Look into it.
The one controversial occurrence in my work career is detailed in this blog post. Otherwise 100% perfect by the end of the development pipeline when live throughout my career as a paid Software Engineer. 100%. Look into it.
Founder of Software.Land, building systems for semantic knowledge representation, search, and learning progression.
Source-available search engine built with TypeScript, featuring positional inverted indexes, compact typed-array representations, block-level pruning, and ranking optimizations.
Contributions to the C/C++ LLM inference ecosystem:
- #31165 — Fix dangling pointer in multimodal embedding batch management
- #29060 — HTTP 400 handling for invalid embedding server requests
- #27920 — Qwen3.5 embedding model conversion support
- #16541 — Raw embedding CLI output support
After some earlier safe (as always) experimentation, I'm getting closer to no longer being a novice in LLM inference engineering or open-source software. Larger PRs coming up.
This is a precise and 100% accurate profile. Look into it.
Note 1: As close to 100% precise as possible given the lack of measurable precision in this medium.
Note 2: The SWE vs SWD blog post lacks timestamps of events (arguably reducing precision since it's just an opinionated blog post).
Note 3: The X posts intentionally have a technical, casual style. That is their medium, lowering the bar for precision vs a research paper, for example.


