I make software,
and I like it when it works.
The honest version of my story — why I started with WordPress, why I left, and why AI engineering is where I'm headed next.
I was eighteen the first time I shipped something to the internet. A friend of my uncle needed a website for his trekking company in Pokhara, and I spent two weeks of summer holiday teaching myself WordPress on a borrowed laptop. He paid me eight thousand rupees, which felt like a fortune, and I was hooked.
For the next three years I built sites — hotels, hospitals, small shops, one boutique publisher who wanted his Gutenberg editor to behave like InDesign. Most of it was unglamorous. None of it was fashionable. But every site had a real client on the other end of it, and every site had to actually work on Monday morning. That's a kind of pressure that turns out to be a good teacher.
The pivot to Python happened almost by accident. A client wanted a scraper for competitor pricing and I'd been reading about BeautifulSoup. I wrote it in two days, billed for four, and realized I liked writing software more than I liked writing templates. Flask came next, then Django, then a couple of years building internal tools and dashboards for small teams.
Then late 2023 happened. The Sora previews, the function-calling announcement, the day Claude actually edited my code without breaking it. I watched the field rearrange itself in real time and knew I was going to spend the next decade somewhere inside it. The question was just where.
I'm not a researcher. I won't publish a paper. But I've spent five years getting software in front of real users, and the thing the AI industry seems to need most right now isn't more clever models — it's people who can wire those models up to the world without everything falling over. That's the job I want.
How I work
I write down what I'm trying to do before I start. I read the code before I change it. I run the thing locally before I claim it's done. When I get stuck I prefer to talk to a person; when no person is available I write the question out as if I were going to ask one, and half the time that solves it.
I'm impatient with abstractions that don't earn their keep, and patient with the unglamorous middle of a project where most of the actual work hides. I take feedback well — I think? — and I try to give it the way I'd want to receive it.
Principles I keep coming back to
Boring tools, interesting problems.
Pick the smallest, most-proven thing that solves the problem. Save your novelty budget for the parts that actually matter.
Make the feedback loop tight.
The faster you can tell whether a change is working, the faster you learn. Most of debugging is shortening the loop.
Read the source.
Documentation rots, Stack Overflow ages, models hallucinate. The source code is what's actually running.
Ship to one real user.
A toy project nobody uses teaches you a tenth of what a real one teaches you. Find the user first.
Eval everything LLM.
Without a test set, you're not iterating on a prompt — you're rearranging deck chairs. Build the eval before the prompt.
Be kind to the next person.
The next person who reads this code is going to be me in six months. Leave them notes. Be generous.
The journey, in dates
-
2026 — now
AI Engineering Intern
Working on retrieval pipelines, evaluation harnesses, and the product surfaces that sit on top of LLM features. Learning how research-grade models become reliable systems.
-
2024 — 2026
Freelance Python developer
Flask & Django apps for small teams — internal dashboards, scraping pipelines, REST APIs. The first taste of writing software that runs unattended and has to keep running.
-
2022 — 2024
WordPress & full-stack web developer
30+ themes and plugins shipped across e-commerce, hospitality, and publishing. Learned to ship on deadline, talk to non-technical clients, and care about Lighthouse scores.
-
2021
Self-taught front-end developer
The borrowed-laptop era. HTML, CSS, and just-enough-JavaScript. Built my first paying website at eighteen.
Where my time goes (last 12 months)
Outside the work
I run early in the morning when Kathmandu is still quiet. I read more than I should and finish less than I'd like. I'm slowly learning to play the sarangi, badly. When I travel I take photos on a film camera I bought from a man in Patan who insisted on drinking tea with me first.