B.Tech CS student building full-stack applications and exploring how production codebases work — including contributing fixes upstream to open-source projects. I work primarily with TypeScript/Next.js on the frontend and Node.js or Python/FastAPI on the backend, with PostgreSQL/Supabase for data.

• Node.js (nodejs/node#65540) — Fixed a bug where fs.cp() and fs.cpSync() (when a filter is used) didn't restore directory timestamps when preserveTimestamps: true was set, even though file timestamps were restored correctly. Open PR; approved by two Node.js maintainers and currently going through CI.

• Dokploy (Dokploy/dokploy#5176) — Fixed two notification bugs: a serverThreshold setting that wasn't being saved for Gotify/Ntfy alerts, and a Teams notification call that wasn't wrapped in error handling and could break the entire notification loop on failure. Open PR; awaiting maintainer review.

• Building Merit AI, a resume-to-job-description matching tool with a Python/FastAPI backend and Next.js frontend
• Contributing bug fixes and regression tests to open-source projects such as Node.js and Dokploy
• Working through DSA problems consistently on LeetCode

A personal finance tracker (work in progress) built with Next.js, TypeScript, Supabase, and PostgreSQL. Uses row-level security to isolate each user's data at the database level. Includes budget tracking, recurring-transaction detection, and an AI assistant using the OpenAI API through the Vercel AI SDK to answer questions about transaction data. Includes a Vitest test suite.

A platform for discovering and voting on AI tools, roughly a Product Hunt for AI. Built with Next.js 16 (App Router / Server Components) and Supabase for authentication, data, and row-level security. Voting uses an atomic PostgreSQL RPC to avoid race conditions during concurrent votes. Includes an admin dashboard for moderating submissions.

A resume-to-job-description matching tool with a Python/FastAPI backend and Next.js frontend. The backend parses PDF/DOCX resumes and uses a deterministic weighted scoring model across skills, experience, tooling, and education rather than letting an LLM generate the score directly. A two-pass LLM step using Groq Llama 3.3 / Gemini cross-checks claimed skills against the resume text to reduce false matches.

272 problems solved on LeetCode, primarily in C++, with 50-day solving streaks in 2025 and 2026.

GitHub · LeetCode