📧 Email • 💼 LinkedIn •

Ever wished you could have a focused conversation with AI without the ChatGPT UI clutter? Building meet.ai - a dedicated meeting interface where you can have deep, recorded conversations with AI that feel more like actual meetings than chat sessions.

The Vision: Instead of scrolling through endless ChatGPT threads, you schedule "meetings" with AI, have structured conversations, and get full transcripts + recordings for reference later. Perfect for brainstorming, debugging complex problems, or working through ideas.

Current progress:

• ✅ Real-time chat interface with OpenAI API integration
• ✅ User authentication with Better Auth (no more sketchy JWT implementations)
• ✅ Conversation recording and storage in PostgreSQL
• ✅ Auto-generated transcripts using Inngest for background processing
• 🚧 Meeting scheduling and session management
• 🚧 Voice-to-text integration for hands-free conversations
• 📅 Next: Analytics dashboard showing conversation patterns and AI usage metrics

Stack: Next.js 14, PostgreSQL, Drizzle ORM, shadcn/ui, Better Auth, OpenAI API, Inngest

Why this stack?

• Next.js 14 – Server actions make API handling clean
• Drizzle ORM – Type-safe queries without the bloat of Prisma
• shadcn/ui – Beautiful components I can actually customize
• Better Auth – Because rolling your own auth in 2025 is asking for trouble
• Inngest – Background jobs for transcript processing without blocking the UI

Sick of carrying 5 different cards just to get into the gym, library, dining hall, and pay for stuff? Building a single app to store all student IDs, passes, and payment cards with NFC support.

The challenge: Most schools use different card systems (RFID, magnetic stripe, barcodes). Reverse-engineering these without violating security protocols is... interesting.

Working on seamless device-to-device screen sharing and file transfer. The goal: share your screen to any device on the same network without installing anything on the receiving device.

Tech exploration: WebRTC, peer-to-peer protocols, low-latency streaming

• Swift & iOS Development – Finally diving into native iOS after years of React/Flutter
• Microservices Architecture – Breaking down monoliths (currently reading "Building Microservices")
• System Design – Prepping for interviews but genuinely fascinated by how Netflix/Uber scale
• Prompt Engineering – LLMs are cool, but making them actually useful is the real challenge
• Drizzle ORM – Loving the type-safety without Prisma's overhead

When I need to ship fast: Next.js + Tailwind + Vercel
When I need it scalable: FastAPI + PostgreSQL + Docker + AWS
When I'm prototyping: Flutter (cross-platform MVP in days)
When I'm learning: Whatever the project needs (recently: Swift, Better Auth, Drizzle)

🤖 Multi-Agent AI Consensus Engine
Ever wonder if an AI's answer is actually correct? I built a system where 5 different LLaMA agents debate and vote on responses. Reduced hallucinations significantly.

The interesting part: Optimized from 8s to 2.5s by running agents in parallel instead of sequentially. Turns out async workflows are magic.

🍱 Oishii – Full Restaurant Platform
Built an entire restaurant management system with online ordering, Stripe payments, and admin dashboard. The hard part wasn't the code—it was understanding PCI compliance and 3D Secure authentication.

What I learned: Financial APIs are terrifying but fascinating. Also, optimizing MySQL queries can 10x your app's speed.

📦 SafeUI – API Wrapper Library
Created a TypeScript package that handles API errors gracefully. Your UI won't crash when the backend decides to take a nap.

1. Ship first, perfect later – Perfect is the enemy of done
2. Test everything – 80% code coverage isn't optional
3. Security isn't a feature, it's a requirement – OAuth, encryption, proper auth flows
4. Performance matters – Users won't wait 8 seconds for a response
5. Learn by building – Reading docs is fine, building is better

Currently: MS in Computer Science @ University of Michigan (2024-2026)
Previously: Worked at a couple startups, built MVPs, shipped production code

I tried to start a company with a friend—we had 45 days to build an MVP for investors. Built a full Flutter app with FastAPI backend while learning OAuth, JWT, and Swift simultaneously. Didn't get the funding, but learned more in those 45 days than in any semester of school.

Now: Interning at PM Accelerator, building AI-powered products, and constantly learning new tech.

I love chatting about:

• 🚀 Building MVPs and shipping fast
• 🤖 AI/ML applications (especially practical ones, not hype)
• 📱 Web development
• 🏗️ System design and scalability challenges
• 💡 Side project ideas and startup stuff

Reach me: dundi@umich.edu | LinkedIn

"The best code is code that ships."