Things I build in my own time to stay hands-on — and to learn what AI can really do in production. I use AI coding agents every day; these projects are where I test what works.
Ruaab
A voice-first CRM for independent real estate brokers in India.
Live
Hold a button and describe a client or a property — in English, Hindi or a mix of both — and the app turns it into an organised record, a follow-up reminder and buyer–property matches.
Works in 4 languages: English, Hindi, Marathi and Gujarati
AI inside
Google Gemini transcribes voice notes, extracts client and property details, rates how likely a client is to close, and drafts follow-up messages.
How I built it
Built using AI as my pair-programmer for development, debugging and hardening.
Why I built it
Most brokers run their business from memory, WhatsApp and paper diaries, because typing into a CRM is a chore nobody keeps up. I wanted capture to cost a few seconds of speaking instead of a form.
What I learned
Local-first mobile apps: everything saves instantly and works offline, then syncs in the background. Getting sync right was harder than building the features.
Turning messy, mixed-language speech into reliable structured data — and designing for the moments when the AI gets it wrong.
Privacy by design: letting brokers collaborate without exposing a client’s personal details, and meeting app-store privacy requirements.
A community app for Swaminarayan sabhas and mandals.
Live
Members check in to sabha and see upcoming birthdays and Atmiya Divas. Coordinators take attendance, follow up with absentees and send WhatsApp wishes in one tap.
Available in 4 languages: English, Hindi, Gujarati and Marathi
How I built it
Built using AI as my pair-programmer, from first screen to store release.
Why I built it
I saw our sabha coordinators spend hours every week on attendance calls and reminders. I wanted to give that time back to them, with an app built around how they already work: on their phones and on WhatsApp.
What I learned
Building one cross-platform app in React Native and Expo, and shipping it to both the App Store and Google Play.
Designing a mobile client that degrades gracefully — when the server changes, the app keeps working instead of crashing.
Building for a real community: four languages, role-based access, and designing around how people actually use it — one-tap WhatsApp wishes and a biometric lock that doesn’t get in the way.
A squad manager for weekly grassroots games — football, cricket, badminton and more.
Beta
Players tap one WhatsApp link to say they’re in and pay the organiser directly by UPI. The organiser hardly has to enter anything — the app works it out from what players already do.
The same 2-screen setup works for 8 sports
How I built it
Designed and built end-to-end using AI — design, code, access rules, tests and the deployment pipeline.
Why I built it
Running a weekly game meant chasing “who’s coming?” and “who’s paid?” on WhatsApp every single week. I wanted an app that figures it out instead.
What I learned
Designing for zero data entry: attendance comes from who said they’re in, games close on their own, and costs split automatically.
Lightweight player identity without sign-up friction, with clear rules on who can do what, and one-time invite links for coaches.