Hi, I am Utsav. I am a senior growth engineer at rtCamp, where I build AI chatbots, personalisation and the martech around them, along with OnePress. Most of it is the same problem in different clothes: get a model to behave usefully against real content, for real users, at a size where the usual advice stops applying.
The interesting part is rarely the model. It is retrieval that returns the right thing, evaluation that catches a regression before a customer does, and a system that stays fast and cheap once it is carrying traffic.
Before this I spent years on large publishing stacks, and I still contribute to WordPress core and Gutenberg, WP-CLI and wp-parsely. Open source is how I learned to write software in public, with people who review it properly.
I write most of it down at blogs.up1512001.com, and every morning at 7 I publish a digest of the dozen things worth reading in AI, engineering and startup funding.
A few things I care about:
AI in production.
A demo that works once is not a feature. The gap between the two is retrieval that returns the right passage rather than a plausible one, evaluation that runs on every change, and knowing what the thing costs per conversation before finance asks.
Most of what I ship sits on top of content somebody already owns, which means the model is the easy half. The hard half is the pipeline underneath it and the honesty to measure whether any of it helped.
Open source.
Patches beat opinions. The fastest way I have found to get good at something is to ship a small fix into a project that thousands of people run, then sit through the review. WordPress gave me that at a scale I could not have bought.
Most of my contributions are unglamorous: performance work, build tooling, editor bugs, CLI ergonomics. That is the job. If you maintain something and want a second pair of eyes, send it over.
Building.
Outside of work I build small products end to end. TypeScript everywhere, Cloudflare underneath, one person on the whole stack. Nothing here has a growth team, so every feature has to earn its keep.
The interesting constraint is not the code, it is deciding what not to build. I am getting better at that, slowly, and mostly by getting it wrong first.
Cars.
The 911 is the reason this page moves. Rear engine, flat six, a layout every textbook says should not work, kept alive for sixty years by people who refused to redraw it and instead engineered around it. That is my favourite kind of stubborn.
The three on your screen drive differently. One follows your cursor, one wants nothing to do with you, one holds its line. None of them will drive over the text.
Away from the screen.
Cricket and badminton, whenever there are enough people for a game. Call of Duty when there are not. More web series than I will admit to. And I read about geopolitics well past the point of it being useful to anyone.
I could not find the sports car of my dreams, so I built it myself. Ferry Porsche
Slow is smooth. Smooth is fast.
Ship the small fix. The big rewrite never lands.
The straight line is where you rest, not where you win.