i'm a product engineer who loves chasing the thrill of 0→1. i enjoy turning things around—especially the messy, overlooked ones that quietly dare you to try.

i care deeply about distributed systems - and even more when it intersects with machine learning. when i'm not coding or whiteboarding, you'll probably find me reading non-fiction, writing, or playing badminton.

this is my tiny corner on the internet where i share thoughts, notes, or works in progress. if something resonates - hit me up.

tl;dr

prev

* oversaw product and investor relations at a $10M ARR startup
* somehow ended up winning more product case competitons than hackathons
* made up for it by shipping a vector DB with HNSW indexing on top of Chrome's IndexedDB
* hit front page of Hacker News, and lived to tell the tale
* shipped open-source competitive intelligence infrastructure for teams
* built a tiny SDK that lets you run inference right inside your browser
* also made an ios app that monitors your posture using your airpods motion sensors
* built a toolkit to help engineering managers stop flying blind
* shipped a tool to help teams transform unstructured data into usable datasets
* built chaos engineering toolkit for teams - re-implemented chaos monkey -
* implemented a B-tree-based database from scratch, because why not?
* shipped a platform to streamline micro private equity buyouts - streamlining due diligence
* built an incident management platform designed for scale-ups

now

* currently i'm heads down building subsignal - helping VCs track and reconnect with startups they passed on
* it's part utility, part ubiquity, and fully an excuse to avoid building a UI
* i also work at Couchbase, owning all things model serving and inference
* shipped 3 out of 5 major undertakings in the last 2 years - from unstructured data service to model serving and inference
* outside of that, i'm trying to learn about distribution - how to get people to care
* other than that, i'm building my writing muscle, reading more consistently, and trying not to burn out
* somehow, i've managed to gain a rep - someone who is intense, takes extreme ownership, and holds the bar high (and then raises it slightly out of reach)
* one who actively seeks conflicts. ask a lot of tough questions, and move like it's your name on the cap table
* not everyone loves that. but things ship, and the bar stays high. maybe that's the whole job
* currently, i'm trying to get better at the less-talked-about half of the job - managing up, and collaborating wide

“browser based ml inference with openai compliant api. run models like deepseek, llama 3.2, nomicembed, kokorotts, and more directly in the browser without server dependencies.”

— tinylm, browser-based ml inference engine, wizenheimer