AI that solvesreal problems.
I build vision-language pipelines, LLM agents and the full-stack products around them. Final-year AI student in Kathmandu, most recently at Reliance Jio.
Technology matters because it solves real problems, not because it is sophisticated.
I have spent 04 years learning to build, shipped 06 projects in that time, and led a team of 06 on the largest of them.
I am in the final year of a BSc in Computer Science and Artificial Intelligence at The British College in Kathmandu, part of the first undergraduate AI cohort in Nepal. There were no seniors to ask and few local resources, which turned independent learning into the habit I rely on most. Most recently I spent a summer at Reliance Jio in Mumbai, putting vision-language models into production.
A few things worth showing.
Open a panel to read about it. Most of these started as something I wanted to exist rather than something I was asked to build.
A personal finance platform you can talk to in plain language. Led a team of six as project manager and lead developer. Django REST, FastAPI, React and React Native over PostgreSQL and Redis, with a Groq-hosted Llama 3.3 70B assistant reading spending history across 60+ endpoints.
Read the case studyFour things, done properly.
The short version of what I am useful for. Everything here is something I have shipped, not something I have only read about.
LLM and agent systems
RAG pipelines, orchestration and model serving with vLLM. Getting a language model to do something reliable and repeatable, rather than impressive once.
Vision-language models
Qwen VLM inference behind a FastAPI service, computer-vision pre-checks that screen input before it reaches the GPU, and turning reviewed footage into datasets for LoRA and QLoRA fine-tuning.
Full-stack engineering
FastAPI, Django, Laravel and Flask on the back, React, Next.js and Vue on the front. Comfortable owning a product from schema to interface.
Shipping and infrastructure
PostgreSQL and MySQL, Docker, CI/CD, REST APIs and auth. The unglamorous half that decides whether any of the above survives contact with users.
Where the time went.
May - Sep 2026
AI/ML Engineering Intern
Reliance Jio Platforms, JPW Division
Built a video verification pipeline on Qwen vision-language models and FastAPI to replace manual field-installation checks, added a computer-vision liveness check to screen spoofed clips before they cost GPU time, and built the workflow that turns reviewed clips into fine-tuning datasets.
2024 - Present
Freelance Full-Stack Developer
Independent
Design and ship production web products for paying clients, most recently a Laravel and Vue.js commercial storefront with inventory, ordering and REST API integration.
2024 - Present
Co-Founder and Core Member
TBC AI Club
Co-founded a student-led group running workshops on applied deep learning, responsible AI and full-stack development, aimed at real problems rather than tutorials.
2023 - 2025
Web Development and Marketing Associate
Subaru Nepal
Built and maintained the company website and internal systems, ran the social accounts and campaigns, and automated the routine workflows that were eating the team's time.
2022 - 2027
BSc (Hons) Computer Science - Artificial Intelligence
The British College, Kathmandu / UWE Bristol
Part of the first undergraduate AI cohort in Nepal. 91% in year one, 89% in year two, on scholarship after both.
What I keep coming back to.
Himal Acharya
On learning
I would rather understand why a model works than just call it. That habit started when there were no seniors to ask and few local resources, and it is still how I learn anything new.
Let's talk.
I am finishing my degree in June 2027 and open to AI engineering work, internships and freelance builds in the meantime. Kathmandu-based, happy to work remotely.