Back to Portfolio

I build what I design—using AI as my implementation layer.

I don't write code traditionally. Instead, I architect products and use AI-native workflows to ship them. This allows me to compress months of development into days while maintaining deep technical understanding.

How I Work

1

Understand the Problem

Deep dive into user research, market analysis, and persona development to find the real friction points.

2

Architect the Solution

Design data models, user flows, and system architecture. Choosing the right tech (e.g., Firestore vs SQL) for the use case.

3

Direct AI to Implement

Using Cursor, Claude, and Gemini as my development team to implement the architecture I designed.

4

Ship and Iterate

Deploying to real users and iterating based on analytics and feedback. Moving at the speed of thought.

My Product Philosophy

Ship to Learn, Don't Plan to Perfection

I built BillFlow's MVP in 1 day to validate demand. Most PMs would spend 2 weeks writing specs. I learned more from 5 real users in 1 week than I would from 10 stakeholder meetings.

AI is a Tool, Not a Strategy

AI Twin and NoteVault use Gemini API, but AI isn't the product—it's the implementation layer. The product is 'better UX through intelligence.' I focus on user problems, not shiny tech.

Constraints Breed Creativity

A recent EdTech assignment had strict requirements (Grade 3-6 math, 10 minutes, prevent drop-off). I chose the hardest topic (algebra variables) to prove gamification works on difficult concepts.

Metrics > Opinions

For BillFlow, I track invoices per user, retention, and feature adoption. For my EdTech projects, I proposed Mission Completion Rate as the North Star. Data tells you if the product works. Opinions don't.

Build for the Worst Case

BillFlow was designed for shop owners with unreliable internet in rural India. If the product works for them, it works for everyone. Design for your least tech-savvy user.

Technical Depth Without Manual Coding

Most PMs hand off specs to engineers and wait weeks. I compress that cycle. I understand database design, API architecture, and security models because I've architected them myself.

I don't manually write production code—AI does that. But I validate, test, and iterate. This makes me a product manager with zero implementation bottlenecks and the ability to speak the language of engineers natively.