CTO. Educator. Builder of things that should exist but don't.
Currently building Jhilke, a mobile-first OTT platform for Nepal, along with PDFMergely and Karyabhar.
I never fell in love with technology itself. I fell in love with what it can fix.
I grew up in Nepal, where the gap between "a problem everyone lives with" and "a problem someone actually solves" is enormous. Farmers lose crops to diseases that are identifiable from a single photo of a leaf. Patients describe symptoms in everyday words that never make it into clinical records properly. Kids grow up without illustrated stories in their own language. Offices run attendance on paper. Everyone accepts this as normal.
I can't. So my pattern became: notice something broken in the real world, learn whatever it takes (machine learning, mobile apps, LLMs, WebAssembly, doesn't matter) and build the fix. Not a demo. Not a paper. A thing people actually use.
That pattern eventually became a job title. As a CTO I do the same thing at a larger scale: pick the problems worth solving, design systems that won't fall over, and grow the people building them alongside me.
Nepal watches on phones, not televisions, yet our films still chase screens we do not own. Jhilke is a mobile-first OTT platform for Nepali short dramas, movies, and original web series, built from Kathmandu for the way people here actually watch. Launching soon on Android and iOS.
Most project tools are built for engineers, and everyone else suffers politely. Karyabhar is a work tracker my non-technical colleagues can use without training: assign a task, do it, mark it done, see the progress. One loop, no jargon, works on any phone. Stack: React, TypeScript, Supabase, Tailwind CSS. Private repo.
I was uneasy that people upload contracts and medical records to random servers just to merge two PDFs. PDFMergely runs every operation inside your own browser, so files never leave your device. Merge, split, compress, sign, OCR, all free, and it keeps working offline. Stack: WebAssembly, Web Workers, PWA
When I became CTO, I inherited a very ordinary problem: who came to work, when, and what the policy says about it. Instead of buying software that almost fit, I built Hajiri, an attendance, leave, and minute-level time-bank system whose rules engine encodes our actual company policy, wired directly to the biometric device at the door. Stack: NestJS, MongoDB, Next.js, Docker
In rural Nepal, a delayed referral can cost a mother her life. SmartAama helps primary health workers keep longitudinal maternal records, flags risk early, and routes referrals to the right facility. The AI is grounded in clinical guidelines and stays strictly advisory, because the final call must always belong to the health worker. Stack: FastAPI, PostgreSQL, React, LangChain
This was my thesis, and the first time I saw my code matter to someone outside a classroom. A farmer points a phone at a sick maize leaf and the app names the disease, explains the symptoms, and suggests a treatment. The Ministry of ICT, Nepal recognized the work in 2024, but the real reward was watching it get used in a field. Stack: Flutter, TensorFlow, Firebase, FastAPI
I wondered whether AI could tell a child a bedtime story worth keeping. Katha Sajha takes a simple idea, writes an illustrated children's tale around it, and lets you download the whole book as a PDF. It is the most playful thing I have built, and it taught me a lot about making AI feel gentle. Stack: Flask, Generative AI, Tailwind CSS
Building is only half of what I do. I teach as an Academic Tutor at Sunway Innovation Project, working with students on data science and AI. Beyond the classroom, I train working professionals: corporate teams learning to use AI in their business, and government officials learning what it means for public systems. I have delivered six complete AI training programs organized by the Government of Nepal.
I keep coming back to teaching for a selfish reason: explaining something to a room of skeptical civil servants is the fastest way to find out whether you actually understand it yourself.
The question I am living inside right now: at Chhayalaya Films we are exploring a pipeline that carries a story from script to screen, with AI helping wherever it honestly helps and staying out of the way everywhere else. The point is not to replace filmmakers. It is to let a small studio in Kathmandu tell stories at a scale it could never afford before.
More broadly, the most interesting problems live at the edges: where machine learning meets a farmer with a cracked phone screen, where an LLM meets a language it barely saw in training, where "state of the art" meets "must not fail on Monday morning." That edge is where I work, and I mentor engineers along the way, because understanding something well enough to explain it is the only proof you understand it at all.
If you're chewing on a problem like that, something real, something unsolved, something that matters to people who aren't in tech, I'd love to hear about it.