PARTHThe Studio

AI Product Engineer

I build AI products
people rely on.

Most AI demos impress.
Production systems survive.
I build the second.

6Products
shipped
2apps in
app stores
Daysidea →
production
Now showing — ApplySync · live in production
scroll
01 — The Work

Real products. Real users. Real impact.

Shipped to app stores and production — not demos.

“The hardest, most human part isn't writing the code.
It's deciding what to build, and being right.”

02 — The Method

Ask my work.

Don't take my word for it — query the record. Pick a Studio Record or ask anything; every answer shows its reasoning and the sources behind it.

Start a conversation →
studio://recordsknowledge engine
>
— How I think

Most engineers show code.
Here's the part that actually matters.

Tools change every six months. Judgment compounds. This is the loop I run on every product.

01
Problem

Start from what hurts and who feels it — never from the model.

02
Judgment

Is AI even the right tool here? Most ideas die at this question.

03
Architecture

Design the system around the model: boundaries, fallbacks, guardrails.

04
Trade-offs

Decide what to make swappable and what to harden. Buy future speed now.

05
Ship

Direct AI for the volume; own the calls. Days, not months.

06
Review

Watch real usage. Keep what earns its place; cut the rest.

— The Builder

What it actually takes.

One night, shipping a fix the team needed by morning.

11:48 PM

Production bug. A few users locked out.

1:12 AM

Traced it to a race condition in the sync layer.

2:14 AM

Redesigned how state reconciles — cleaner, safer.

4:30 AM

Tests green. Edge cases covered.

5:41 AM

Deploy complete. Logs quiet.

9:02 AM

The team logs in. It just works. Nobody knows.

~/deploy● live
— Field Notes

Lessons, in my own words.

The opinionated stuff — what I rejected, what broke, and what I'd do again.

Prompting isn't engineeringAI

A clever prompt is a demo. The engineering is the system around it — fallbacks, daily budgets, guardrails, a kill switch. That's what survives contact with real users.

The Firebase decisionArchitecture

For ApplySync I chose Firebase over Postgres. It cut deploy from days to hours and kept the early build simple. Postgres would have doubled the complexity for value I didn't need yet — and I can migrate later if scale demands it.

Distribution beat codeLesson

ApplySync worked end to end and launched to almost zero users. The lesson reorganized how I build: shipping isn't the finish line, it's the start of the real problem — getting people to use it.

Governing AI you can't fully trustHealthcare

For I Am Still Alive I assume the model will sometimes be wrong. So: de-identify before any call, audit every request, cap usage, and keep one switch that halts all AI instantly. Safety over uptime.

Why one swappable AI layerArchitecture

On the conference platform every model sits behind a single interface. When a better/cheaper model appears, it's a one-line swap — not a refactor across 35 data models.

Parth Kothawade
03 — About

I don't build demos.
I build products.

I care about systems that work. I care about users who trust them. I care about impact that compounds. I'm an AI-native builder from Chalisgaon, Maharashtra — I design the architecture, direct the AI, and own the outcome end to end.

Parth
Foundations

Learned by building — apps, full-stack, the hard way.

Production

Real systems: healthcare AI, events, internal tools.

AI-native

6 products shipped by directing AI end to end.

Now

Building systems people rely on, from Chalisgaon, Maharashtra.

— The Journey

Becoming the builder.

Not a résumé — a path. What I built, in order.

2022
Joined VIT VelloreElectronics & Communication Engineering
·
Fundamentals firstDSA · competitive programming · problem-solving
·
Learned to buildWeb development · Flutter · Firebase · React
·
Humanity Founders Hub · US startupMobile Engineer — shipped production Flutter apps
·
I Am Still Alive · US healthcareProduct Developer — architecture & product decisions
·
Built ApplySyncFounder · AI Product Engineer — full-stack + AI, solo
2026
GraduatedB.Tech · Electronics & Communication
now
Shipping at I Am Still Alive & ApplySyncFrom Chalisgaon, Maharashtra — idea to production, end to end
— The Missions

Where I've built.

Three teams. Three roles. What I actually shipped.

Mission 01

Humanity Founders Hub

US startupMobile Engineer

Shipped production Flutter apps end to end — authentication, REST integrations, performance tuning and real production debugging across Android & iOS.

BuiltFlutterFirebaseREST APIsIn-app purchases
Mission 02

I Am Still Alive

US healthcare platformProduct Developer● Current

Owned architecture and product decisions on an oncology-support platform, with AI-assisted development — working directly with CEO Danielle Bellini.

BuiltFull-stackGoverned AIArchitectureProduct decisions
Mission 03

ApplySync

My startupFounder · AI Product Engineer● Current

An AI SaaS that discovers jobs, rewrites résumés per ATS, applies through Gmail and auto-fills forms via a Chrome extension — designed and built solo, front to back.

BuiltReactNodeFirebaseOpenAI · NIMChrome MV3Razorpay
— The Stack

What I can build.

Not a badge wall — the capabilities behind the products.

Frontend Engineering

Interfaces people actually use.

ReactNext.jsFlutterTailwind

Backend Systems

APIs, data, auth, real-time.

NodeExpressFirebaseSupabaseDjangoPrisma

AI & LLMs

Systems around the model — not just prompts.

OpenAINVIDIA NIMRAG / embeddingsPrompt engineering

Infrastructure

Ship it and keep it running.

VercelRailwayGitHub ActionsCloudflare R2

Product & Automation

Turning a demo into a product.

Chrome MV3Telegram botsCron jobsRazorpay
— The Record

Proof over promises.

Milestones, open source, and what I'm building right now.

Milestones

  • 6 production systems shipped
  • 2 apps live — App Store & Google Play
  • AI products running in production
  • A Chrome extension across 46+ sites
  • A governed healthcare-AI platform
  • Founder of an AI SaaS

Featured repositories

All repos on GitHub →

Currently

  • Building ApplySync
  • Learning agentic systems
  • Research production AI architecture
  • Reading Designing Data-Intensive Applications