Full stack developer specializing in React, FastAPI and Flask. I've shipped production features on a live trading platform, built an ML app in 24 hours flat, and I write code that's meant to be read by the next person, not just the compiler.
I'm Harshwardhan, a 3rd year Computer Science student specializing in AI/ML, and a developer who'd rather ship something real than talk about shipping something real.
My internship at a fintech startup had me building a real-time trading platform , React on the front, Flask and WebSockets underneath, wired into the Angel One SmartAPI.
That's where I learned the difference between code that works in a demo, and code that has to stay up while money moves through it.
Outside of client work, I compete LungAI, an ML powered lung cancer detection tool with Grad CAM visual explanations, was built in a single 24-hour hackathon with my team. I've also sent a fix upstream to TheAlgorithms/Python, one of GitHub's largest open-source repositories.
These days I split my time between client projects and building tools for small businesses that need a real web presence not a Wix template with their logo pasted on.
Each one taught me something the last one couldn't a trading dashboard is not a diagnostic tool, and neither is a hospital site.
Production feature work on a live stock trading web app during my internship live price feeds over WebSockets, order execution flows, and a React dashboard that can't afford to lag.
View repositoryA full hospital web presence built to show what a client's site could look like beyond a template booking flows, department pages, doctor profiles, and a design system with actual restraint.
View liveAn artisan coffee roastery web application featuring interactive menus, dynamic origin stories, and a warm, editorial design system crafted for high user engagement.
View liveBuilt in 24 hours with team Bibi Bytes for PCU Ideathon 3.0 EfficientNetB0 under the hood, Grad CAM heatmaps so doctors can see what the model sees, FastAPI serving it all.
Case studySubmitted a fix for broken doctests in xgboost_classifier.py to one of GitHub's most-starred educational repositories. Approved by 1 reviewer, awaiting a second approval for merge.
View PR #14224Red Hat systems training and a run of Coursera specializations, done alongside coursework.
Not "build me a website" what's it for, who's it convincing, and what happens if it doesn't work. Fifteen minutes here saves a week later.
Staged previews as I go, not one big reveal at the deadline. You can redirect early, when it's cheap to.
Live, responsive, and documented well enough that you're not locked to me for every future tweak.
I take on a small number of client builds at a time reach out and let's see if it's a fit.