AI engineer & full-stack developer — India
IbuildAI productsthatruninproduction.
(01) — About
Complete products, not demos — nothing is real until it's wired to an actual database, an actual API, and actual users.
- Based in
- Punjab, India · IST
- Shipping publicly
- since 2023 · 50 repos
- Best at
- AI agents, applied ML, full-stack products
- Also do
- Voice agents, dev tooling, React Native
- Languages
- Python, TypeScript, JavaScript, SQL
- Elsewhere
- Parthav AI on YouTube
I'm Parthav — an AI engineer and full-stack developer from Punjab, India. Scroll my GitHub and you'll see almost every idea shipped as a pair: a React or TypeScript frontend with a Python or Django backend behind it, because a feature isn't finished until real users can touch it.
I came up through classic full-stack work — React, React Native, Python, Django — and learned ML the practical way, by building with it. That path runs from early data-analysis notebooks and a Spark house-price model, through production web apps like EzPrints and a full LMS, to NTF Travel, a live IATA-certified travel booking platform. In 2025 I went on an applied-AI sprint and shipped a string of end-to-end products: Cardio Care AI, Crime Radar AI, AgriVision, Job Sphere AI, Moodify and more.
These days my work centers on AI agents and developer tooling in TypeScript and Python. I built docgen — an open-source CLI that documents any codebase in about a minute using Claude, OpenAI, or Gemini — and I'm deep in agent work: voice agents, automation pipelines that quietly save clients hours every week, and products like REMLY-AI and OrderKaro that are in active development right now.
Away from the editor I run Parthav AI on YouTube — honest, hype-free takes on the tools I actually use, and live streams where I build real projects on camera. I'm based in Punjab (IST) and I reply fast. If you're building something with AI, let's talk.
Recent builds
04 shippedShipped as frontend + backend pairs — real users touched every one. Hover a card to inspect.
(02) — Selected work
A library of things I've shipped.
Client platforms, AI products and open source. Drag, scroll or swipe the wall — it loops endlessly through the full catalogue; open the live ones.
17 shown · 50+ repos on GitHub · new work every month
(03) — What clients say
Shipped, and trusted.
“Parthav rebuilt our trading dashboard and cut p95 from 3.2s to 240ms. Traders stopped complaining — that never happens.”
“Shipped our carbon-ledger MVP in six weeks and it just worked in production. MRR is up 38% since launch.”
“Quoting went from four hours to under forty minutes. He understood the operation, not just the code.”
“The store feels expensive and it converts — AOV up 22%. Worth every rupee.”
“Parthav rebuilt our trading dashboard and cut p95 from 3.2s to 240ms. Traders stopped complaining — that never happens.”
“Shipped our carbon-ledger MVP in six weeks and it just worked in production. MRR is up 38% since launch.”
“Quoting went from four hours to under forty minutes. He understood the operation, not just the code.”
“The store feels expensive and it converts — AOV up 22%. Worth every rupee.”
“No-shows dropped 41% once the reminder agent went live. He builds things that move numbers.”
“Prototype to 12k users on infra Parthav set up. It hasn't fallen over once.”
“Agents open their day with briefs, not phone numbers. That was the whole point — and he got it.”
“10M events a day and the dashboards are still instant. Genuinely impressive engineering.”
“No-shows dropped 41% once the reminder agent went live. He builds things that move numbers.”
“Prototype to 12k users on infra Parthav set up. It hasn't fallen over once.”
“Agents open their day with briefs, not phone numbers. That was the whole point — and he got it.”
“10M events a day and the dashboards are still instant. Genuinely impressive engineering.”
(04) — What I build
Complete products,
not demos.
(06) — How I work
From a messy workflow to a system that ships.
Every engagement runs the same five beats — fast where it should be fast, careful where production demands it. Hover a step to see what gets built.
Find the loop worth killing
Discovery · scopingWe map the real workflow together — the inbox that gets copy-pasted, the report built by hand every Monday. I don't automate what shouldn't exist; I find the one loop that quietly eats hours and agree on what 'done' actually looks like.
A working slice in a week
PrototypeNo decks, no mockups you can't click. Within days there's a thin end-to-end slice wired to real data and a real API, so you can feel the thing before we commit to the whole thing.
Build it for production
EngineeringThis is where most AI demos die and where I live — auth, queues, data models, retrieval that answers from your data, and an eval harness that gates every deploy. The unglamorous layer that decides whether it survives real users.
Ship with a human in the loop
DeployIt goes live behind a human check first. We watch it in the wild on real traffic, catch the weird edge cases, and only widen the autonomy once the numbers say it's earned.
Operate, measure, scale
OperateLatency, cost per run, correctness — tracked, not guessed. I stay on to keep it fast and cheap as usage grows, and to turn the next loop into the next build.
04 — THE RECEIPTS
Shipping in the open.
Fifty public repositories and counting since 2023 — backends, AI products and CLIs, pushed continuously. Not slideware. Code you can clone.
0
Public repos
0
Stars earned
0
Contributions
0
Followers
981 contributions · last 52 weeks
Selected repositories
docgen
Docs for any codebase in 60s — multi-LLM (Claude / OpenAI / Gemini) Node CLI. Architecture overviews, API routes, dependency graphs. MIT.
CRIME-RADAR-AI-BACKEND
Django + TensorFlow backend for Crime Radar AI — public-safety analytics served over a real REST API, paired with its React frontend.
NTF-Travel
Full-stack travel platform live at ntftravel.com — bookings, itineraries and payments wired end-to-end. TypeScript + Next.js.
EZPRINTS-BACKEND
Backend of EzPrints, an automated document-printing product shipped as a three-repo system — backend, storefront and admin.
ShopSavy
E-commerce build with cart, checkout and an order pipeline — JavaScript frontend against a REST backend.
CARDIO-CARE-AI-BACKEND
Heart-disease prediction served over an API — model training, inference endpoints and a JavaScript client. Django + scikit-learn.
06 — THE HANDSHAKE
LET’SBUILD.
Got a product that needs to actually ship — wired to a real database, a real API and real users? Send the brief. I reply within 2–4 hours.
parthavsabrwal@gmail.com