Utsab Adhikari

  Full-stack engineer — London

I build and ship complete products — API to deploy, solo.

Three years building production apps with FastAPI, React/Next.js, PostgreSQL and LLM APIs — backend fundamentals through to a deployed URL, without a team to hand off to.

  • FastAPI
  • React
  • Next.js
  • PostgreSQL
  • Supabase
  • LLM APIs

System shape

A typical system: React or Next.js frontend talks to a FastAPI or Node API over HTTP and WebSocket; the API reads and writes PostgreSQL on Supabase and calls LLM APIs such as Claude for inference.

01 — Selected work

Four products, built and shipped end to end.

Two with full case studies. Every link here goes to a live demo or the source — nothing that 404s.

Case study

AI LMS

Live

Generic online courses don't adapt to where a learner actually is.

A learning platform where Claude generates the curriculum and quizzes, mastery updates from quiz performance, and a realtime AI tutor answers over WebSocket.

Key decision

Quiz submissions update an exponential-moving-average mastery score per topic, so the next content targets weak spots instead of starting over.

  • React
  • TypeScript
  • Vite
  • FastAPI
  • Anthropic Claude
  • Supabase
  • WebSocket

Case study

VisaTrack

Live demo soon

People on UK visas need to keep their status and the documents that prove it in one reliable place.

A UK visa tracker on Next.js and Supabase — sign-in, document storage, and status tracking in one app.

Key decision

One managed backend: Supabase provides auth, Postgres and document storage, so there are no separate services to run or secure for a solo-built product.

  • Next.js
  • TypeScript
  • Supabase
  • PostgreSQL

Also shipped

AI Influencer Generator

Keeping an AI persona visually consistent across many generated images is hard.

Image generation runs as Celery/Redis background jobs writing to S3-compatible storage, so the UI stays responsive during slow generations.

  • Next.js
  • TypeScript
  • FastAPI
  • Celery
  • Redis
  • PostgreSQL
  • S3

CiteSnap

Formatting academic references by hand is slow and error-prone.

Fully client-side with history persisted locally: zero backend, instant results, and free to run.

  • JavaScript
  • HTML
  • CSS

02 — How I work

A few things that are true of every build.

01

I ship end to end

Schema and API through to a deployed URL, done myself. The projects above are solo builds, not team hand-offs.

02

AI-augmented, and I build the tooling

I work with Claude Code daily and build my own agent tools — like an LLM job pipeline that scrapes, scores with the Claude API, and automates applications with Playwright.

autoapply-agent
03

Backend fundamentals first

Typed APIs, Postgres schema design, background jobs and auth before UI polish — like the Celery/Redis job pipeline behind AI Influencer Generator.

03 — About

Full-stack engineer, London.

I build complete products end to end — the API, the database, the frontend, and the deploy. I care most about the parts that decide whether software actually works in production: data modelling, background jobs, auth, and clear failure modes.

I've worked at Pagoda Labs and We Media and hold an MSc in Software Engineering. My current focus is AI-forward products — building with FastAPI, React/Next.js and the Claude API.

Based in
London, UK
Education
MSc Software Engineering
Previously
Pagoda Labs · We Media
Focus
AI-forward product engineering

04 — Contact

Two ways to start a conversation.

Hiring for a role

Full-time or contract. The fastest read on how I work is the case studies above, then my CV.

LinkedIn soon Download CV (PDF) soon GitHub

Have a project

Need something built end to end — an API, a web app, an AI feature? Email me a sentence about it, or reach me on Upwork.

Upwork profile soon Email me