I am a full stack developer from Vadodara, currently doing an M.Tech in Artificial Intelligence and Data Science at Parul University. Before that I spent four months at Mamo Technolabs writing production React and Node.
Most of what I build is MERN: real time comment trees, booking platforms, campus APIs. My dissertation asks a narrower question, which is what an AI coding agent does after it gets something wrong.
Currently M.Tech dissertation on failure recovery in agentic coding systems
Building Recovery Bench, a benchmark for self-repair in AI coding agents
Learning Distributed systems, LLM evaluation, benchmark design
Based in Vadodara, Gujarat, India
A Self-Reflection-Based Failure Recovery Framework for Agentic AI Coding Systems M.Tech Dissertation · Parul University · 2026
Agentic coding systems are measured on whether they solve a task, not on what they do when they get it wrong. I built a benchmark that seeds controlled failures into coding tasks, then measures whether an agent can detect, diagnose and repair 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.
| Result | |
|---|---|
| Self-reflection prompting | 3 / 3 recovered across the seeded failure set |
| Baseline retry | 0 / 4 no recovery without an explicit reflection step |
| Test suite | 433 tests across a Bun and TypeScript workspace |
Frontend
Backend & Data
Tooling
| Project | What is interesting about it | Live |
|---|---|---|
| Recovery Bench | Benchmark harness behind the dissertation. 433 tests, Bun workspace. | dashboard |
| nested-comments | Live comment trees where a child can arrive before its parent, so orphans are buffered and spliced in later. | demo |
| Hunger Hive | Storefront, admin dashboard and Express/MongoDB API over one data model. | demo |
| Find My Driver | Three roles over a single Express API with role based access control. | demo |
| EVM Reality Check | Each common misconception paired with a simulation you can step through. | demo |
| CampusIQ 360 | Campus management REST API. Attendance, assignments, timetables, events. | API only |
Open to internships, research collaboration and freelance work. Email is fastest: nakullagad084@gmail.com