All work
Everything I have built, from production AI systems to developer tools to full stack apps. Click any card to read the full story.
20 projects total
A key value store built from scratch in Python to understand what Redis is actually doing internally. Hand written RESP protocol parser, lazy TTL expiration, crash safe persistence via an append only log, and thread per connection concurrency, all without a single external database library.
Read full story →
An autonomous AI agent that detects crypto volume and price anomalies, then explains the catalyst behind them using real news, LangGraph, and Gemini 2.5 Flash.
A lightweight LSTM neural network trained on 13 years of Bitcoin minute-level data to predict the next hour's closing price, served via FastAPI.
BotSpark is a multi-tenant AI chatbot SaaS. A business connects its own data and brand, deploys a custom AI agent in minutes, and it answers customer questions 24/7 instead of a team manually replying. Built with Next.js, NestJS, and MongoDB, with multi-provider LLM fallback so a single provider outage doesn't take a tenant's bot offline.
AI Code Reviewer is an automated system that reviews every GitHub PR. It posts structured feedback on bugs, security, and performance directly in the PR thread. Built with FastAPI, Redis for task queuing, and an advanced AI engine for analysis.
SEO is the lifeblood of any web product, but most audit tools are either too generic or prohibitively expensive for solo developers. They hand you a list of checkboxes— missing alt tags, slow LCP—without any real context about why it matters or what to actually fix first.
ContentAgent is a multi-agent system built with LangGraph. It breaks down technical writing into a pipeline: Researcher -> Outline Creator -> Writer -> Critic -> Formatter. It produces ready-to-publish technical blogs from a single prompt.
QuickPU is a result portal for University of Punjab students. Instead of digging through 350+ different department result URLs to find one roll number, students search once and get their result. On result day it handled 850 concurrent users and 4,100+ API hits, deployed on EC2 with FastAPI and MongoDB.
Journaling is a powerful tool for mental clarity, but many people struggle to see the big picture of their emotional health. MyMindfulJournal was built to turn simple daily entries into a meaningful map of your mood. It’s not a therapist; it’s a mirror that reflects your patterns through data.
Articles on building AI systems, web apps, automations. No fluff. Code you can actually use.
Tap to view →
We spend half our lives inside messaging apps. Why open a separate app for your notes, another for your calendar, and another to ask an AI a question? I wanted to build an assistant that lived where I already was: WhatsApp.
React changed how we think about state with its unidirectional data flow and reactive updates. I wanted that same mental model in Python—not necessarily for the browser, but for any Python logic. Whether you're building a CLI, a background service, or a lightweight dashboard, managing state shouldn't involve a mess of callbacks and global variables.
Redux is too heavy for small projects. Context API can lead to unnecessary re-renders if not handled perfectly. I wanted a library that felt as simple as useState but worked across the entire application.
Researchers often work with hundreds of documents, reports, PDFs, and academic papers at the same time. Reading everything manually takes weeks, and finding specific information across multiple files becomes increasingly difficult as datasets grow.
Good UI shouldn't be hard to find. While libraries like Radix and Shadcn provide the primitives, I often found myself rebuilding the same Hero Section or Feature Grid patterns across different projects. Ui-Hatch is my personal vault of these high-level components.
You are looking at it. Built to be fast, clean, and honest about what I do.
Every time I started a new NestJS project, I spent the first 30 minutes doing the same thing: setting up MongoDB connections, configuring JWT authentication, adding Swagger for API docs, and organizing the folder structure. It was boring, repetitive, and error-prone.
Running npx create-next-app is just the beginning. Most developers then spend the next 20 minutes installing shadcn/ui, configuring their globals.css for Tailwind v4, setting up a consistent folder structure, and adding their favorite utility functions.
As a developer, I find myself constantly switching tabs: searching for an error on Stack Overflow, checking a repo on GitHub, and looking up a package version on npm. I wanted a single command center for all developer knowledge.
Businesses want AI assistants, but building one usually requires developers, infrastructure, document processing pipelines, prompt engineering, and ongoing maintenance.
Ask me anything 👋