M.Tech in AI & Data Science at Parul University
Nakul Lagad
Full stack developer and M.Tech researcher in Vadodara. I ship MERN applications, and my dissertation asks a narrower question: what does an AI coding agent do after it gets something wrong?
About
I am a full stack developer from Vadodara, currently doing an M.Tech in AI and Data Science at Parul University. Before that I spent four months at Mamo Technolabs writing production React and Node, which is where I learned the difference between code that works and code that holds up.
Most of what I build is MERN. Real time comment trees, booking platforms, campus APIs. The projects below are the ones where something was genuinely difficult, not the ones with the longest feature lists.
My dissertation measures whether an AI coding agent can diagnose its own bad patch, rather than whether a retry happens to work.
- 6 projects shipped
- 433 tests written
- M.Tech researcher
Selected work
All repositoriesO(1)
orphan reattachment
nested-comments
Real time threaded comments over WebSocket
Comment trees that stream in live. The hard part is out of order arrival: a child can reach the client before its parent exists, so the tree holds orphans and splices them in once the parent lands.
React 19 TypeScript Zustand ws
Research
M.Tech Dissertation · 2026
When an AI coding agent fails, can it fix its own mistake?
Agentic coding systems are measured on whether they solve a task, not on what they do when they get it wrong. This work builds a benchmark that seeds controlled failures into coding tasks, then measures whether an agent can detect, diagnose and recover from its own bad patch. The framework separates recovery from retry: re-running a prompt is not the same as reasoning about why the first attempt failed.
3 / 3
Self-reflection prompting
recovered across the seeded failure set
0 / 4
Baseline retry
no recovery without an explicit reflection step
A Self-Reflection-Based Failure Recovery Framework for Agentic AI Coding Systems
Tools I work with
Frontend
- React
- Next.js
- TypeScript
- Tailwind CSS
- Zustand
- Framer Motion
Backend
- Node.js
- Express
- MongoDB
- Mongoose
- WebSocket
- Redis
- JWT
Tooling
- Git
- Docker
- GitHub Actions
- Vite
- Bun
- Vercel
- Render
Research
- Agentic AI
- LLM evaluation
- Benchmark design
- Python
Background
Experience
Software Developer
Mamo Technolabs · Vadodara, India
Dec 2024 - Mar 2025
- Built and maintained full stack features across React, Next.js, TypeScript, Node.js and Express.
- Designed REST endpoints for authentication, order management and payment flows.
- Reduced response times on read-heavy endpoints by introducing Redis caching and moving blocking work off the request path.
- Containerised services with Docker and automated builds through GitHub Actions.
Education
M.Tech, Artificial Intelligence & Data Science
currentParul Institute of Engineering & Technology, Parul University
2025 - present · Vadodara, India
B.Tech, Computer Science
Parul University
2021 - 2025 · Vadodara, India
Let's build something.
Open to internships, research collaboration and freelance work. The fastest way to reach me is email.
- Phone
- +91 93283 21950
- Location
- Vadodara, Gujarat, India