Applied AI & Full-Stack Developer
LLM Evaluation · AI Coding Agents · Production Web & Mobile Systems
- 🚀 Full-Stack Developer Intern at Anytime Diesel, building production web, PWA, Android, and iOS applications.
- 🤖 Previously worked on LLM and AI coding-agent evaluation, including benchmark tasks, deterministic testing, Docker environments, and model failure analysis.
- 🔬 Former INSA Summer Research Fellow at IIT Ropar, working on REST APIs, recommendation-system evaluation, and frontend performance.
- 🎓 B.Tech in Computer Science and Engineering from IIIT Manipur (2022–2026).
- 🧠 Strong foundation in Data Structures, Algorithms, OOP, REST APIs, SQL/NoSQL databases, Git, and CI/CD.
- 🌱 Interested in Applied AI, AI-enabled product engineering, full-stack development, and developer tools.
| Area | Technologies and Focus |
|---|---|
| Applied AI & Evaluation | LLM evaluation, AI coding agents, benchmark design, oracle solutions, failure analysis, and reproducible testing |
| Product Engineering | React, TypeScript, Node.js, Express, REST APIs, MySQL, MongoDB, Prisma, and responsive interfaces |
| Quality & Delivery | Pytest, shell testing, Playwright, Docker, GitHub Actions, CI/CD, AWS, Azure, and Linux |
06 July 2026 – Present
- Building Anytime Diesel Workforce, a production HRMS available across web, PWA, Android, and iOS.
- Working with React, TypeScript, Node.js, Express, Prisma, and MySQL.
- Delivered role-based access control, GPS and face-enabled attendance, leave management, asset management, expense workflows, notifications, and audit trails.
- Built an Inside Sales Dashboard for OMS claims, targets, approvals, customer ownership, and reporting.
- Implemented JWT authentication and Dockerized deployment.
- Reduced avoidable OMS API traffic using persistent synchronization, local-first reads, targeted refreshes, freshness windows, and background cooldowns.
- Created and reviewed hard coding tasks for evaluating LLMs and AI coding agents.
- Designed realistic implementation, debugging, build, CLI, and dependency-management workflows.
- Built Docker-based task environments, reference solutions, and deterministic
pytestand shell verification suites. - Reviewed AI-generated solutions for correctness, completeness, instruction following, regressions, and edge-case handling.
- Identified specification-test mismatches, weak coverage, nondeterministic tests, Docker failures, and invalid oracle behaviour.
- Documented model failure modes, reviewer feedback, and task-quality improvements.
19 May 2025 – 13 July 2025
Project: LearnFlow
Mentor: Dr. Sudarshan Iyengar, Associate Professor, CSE
- Designed REST API integrations between a React frontend and MongoDB backend.
- Reduced average API response time by approximately 20% through targeted query optimisation.
- Co-built a Python synthetic-learner simulator using Pandas and NumPy to evaluate recommendation-engine accuracy.
- Improved model precision by approximately 18%.
- Re-architected the React and Bootstrap frontend using reusable components and responsive layouts.
- Improved usability and reduced page-load time.
A role-based campus entry and visitor-management platform built using the MERN stack.
Key features:
- Secure authentication and role-based access
- Visitor approval workflows
- QR-based entry verification
- Responsive administrative dashboards
An automated content pipeline that collects horoscope data, generates daily videos, and publishes them to YouTube.
Technologies: Python, Selenium, FFmpeg, and GitHub Actions
- MongoDB Associate Developer — MongoDB University
- Data Analytics with Python — NPTEL, IIT Roorkee
I'm open to opportunities in Applied AI, LLM Evaluation, AI Agent Engineering, Full-Stack Development, and Software Engineering.