twitter • linkedin • github • medium • email

• Recast - Agentic video recorder and editor

• Docvia - framework agnostic documentation compiler

• GlyphX - A fast, powerfull LaTeX editor that Overleaf should have been

• College Ecosystem - Next.js/TS platform; 1M+ visits; realtime polls, rankings, communities.

• Orbit - browser-first pdf utility application

• Mailing system - A modern, type-safe & portable email service built with Typescript, React Email, and Nodemailer.

• Markdown Editor - plugin-based Markdown Editor with React + TS + shadcn UI / Tailwind css ; rendering speed +35%; 300+ weekly downloads.

• Rich Text Editor - A lightweight, customizable React Rich Text Editor built on TipTap

• Remark Plugins - A collection of remark plugins for markdown processing.

• Custom Domain Management SDK : a framework-agnostic SDK for multi-tenant domain onboarding using strict lifecycle state machine

• Muse : AI-powered fashion discovery

• NexoNauts.com | Live : Ecosystem for developers (Next.js, Shadcn UI, MongoDB)_

• Bruno - Opensource IDE For Exploring and Testing API's (lightweight alternative to Postman/Insomnia)
• Optexity - Build custom browser agents with AI-powered automation

• I build from zero. Whether it's frontend, backend, full-stack applications, or AI-powered experiences, I work across the entire development lifecycle. From UI/UX to deployment to user feedback, I care less about technology debates and more about delivering results that people love using.
• Domain knowledge of Full-stack AI apps development, scalable systems design, and devops/cloud computing.

SDE - Full Stack | Mar 2026 - Jun 2026

ViralLens is a digital marketing company. I worked on its tech side building software for a US law firm's class action practice, a short full-time run of about four months where I owned a lot of ground across local-first tooling, AI pipelines, and observability.

• Built a local-first markdown engine and rich text editor that runs entirely in the browser on IndexedDB. Full offline storage, multi-document handling with tabbed editing, and PDF export. It can lean on up to 80% of the device's storage quota when it needs to, so notes and prompts stay on the machine and never touch a server.
• Built a DOCX MCP that lets multiple AI agents edit and collaborate on the same document with low-level API control. It exposes tools for the parts a real legal document needs: create, search, and update sections, a table of contents, and a table of authorities.
• Built the AI agent pipelines behind the product: content generation, content analysis, and fact verification, with per-section scoring and highlighting so you can see exactly what held up and what didn't.
• Built market and legal analysis tools that score stock performance and estimated damages, then predict a market reaction score for how a given class action might land, alongside a good amount of legal text processing.
• Set up the observability layer: a unified analytics platform tracking how long processes run and what they cost, per-application error and failure tracking with Better Stack, and PostHog on client projects to see how people actually used what we shipped.
• Wrote the internal documentation and refactored the codebase early on to make the rest of this possible.

Software Developer | Dec 2022 - Jun 2024

I joined right as ChatGPT went mainstream, when the whole plan was to ship AI tools fast and see what stuck. Started as a five month intern and stayed on part time, mostly on the frontend but picking up auth, deployments, and backend whenever the small team needed it.

• The founder wanted a separate app for every idea he could think of: LinkedIn bio generator, summarizer, resume writer, "turn this into that", and a long list beyond. I first moved the React codebase to Vite to speed up the dev loop, then migrated everything to Next.js once we were juggling around 20 of these tools and the old setup turned slow and buggy.
• Twenty hardcoded tools was never going to scale, and a separate endpoint per tool made no sense. So I built a low-code, drag-and-drop tool builder. You lay out a form (text, textarea, select, multi-select), each field gets a unique id, and you write one prompt that references those ids by mention. At runtime we swap the mentions for the user's real input, send it through the OpenAI call, and stream back the text or image. One engine replaced twenty apps, and anyone on the team could ship a new tool without an engineer.
• Owned deployments end to end. We started on Azure shipping through GitHub Actions, then I moved us to GCP with Docker and their container registry, which made releases far more predictable.
• Migrated authentication from Amazon Cognito to NextAuth. The hard part was sharing a session securely across subdomains, which I solved with correctly scoped, secure cookies.
• When we pivoted to news and financial analytics, I worked the document-analysis side with LangChain and web search, turning results into charts on a multi-tenant setup.
• Reworked the UI and UX as we grew to roughly 10,000 signed-up users, with a share of them converting to paid.

Frontend Engineer | Mar 2024 - Jun 2024

KoinX builds crypto tax software, the tool people across different countries use to work out and report what they owe on their crypto. I came in as a frontend engineer for a three month internship and worked across both sides of the product.

• Started on the consumer dashboard, shipping new pages, charts, and graphs, then moved onto an internal tool used by chartered accountants along with a few team-only features.
• On my own, without being asked, I branched one of our production apps and migrated it from Create React App to Vite for a much better developer experience. My lead liked the result and asked me to do the rest, so I migrated all five to six production CRA repos over. The trickiest part was SVG handling, which I sorted out with the right Vite plugins.
• Built new components for the internal UI component library so the interface stayed consistent across our apps.
• Our marketing site was expanding to multiple languages and countries, but it hydrated everything on the client and pulled all content from a single JSON file, so it kept getting slower. I moved rendering to build time and server-side in Next.js, and made the content imports tree-shakable so a page only loads what it needs.
• With no extra libraries, just lazy loading, dynamic imports, and runtime code splitting, I took the site's PageSpeed performance score from around 60 to 88 and improved Core Web Vitals across the board.

Languages : JavaScript, TypeScript, Go, Python, Rust, Php, C++

Frontend & Frameworks : React.js, Next.js, Svelte 5, Sveltkit,Tauri, Vite, TailwindCSS, Shadcn UI, SASS, Turbopack

Backend: Node.js, Express.js,Hono, FastAPI, Go Fiber, REST APIs, Microservices

Databases: PostgreSQL, MongoDB, Redis, Firebase, Supabase

Cloud/DevOps: Docker, Kubernetes, AWS, GCP, Azure, GitHub Actions, Vercel, CI/CD

Other: Algorithms & Data Structures (150+ LeetCode), System Design, Performance Optimization, UI/UX

⚡ Recent Activity:

1. ℹ️ Assigned PR #17 in kanakkholwal/glyphtex