A curated collection of my work spanning full-stack development, artificial intelligence, and specialized engineering tools.
An On-Premises Intelligent Document Processing & Field Extraction platform. Built for high-concurrency extraction of structured data from complex industrial documents. Features live page-by-page incremental processing, bidirectional canvas hit-testing, and hallucination zeroing with confidence fusion.
A local-first semantic search engine entirely offline and built in Rust. Features 3 parallel search arms (Exact Keyword, Fuzzy Match, Semantic Intent) and multi-modal data support for notes, codebases, Wikipedia, and PDFs. Includes a full-stack Web UI and RAG integration.
A modern, full-stack automation bridge that monitors Instagram interactions to trigger automated reel capture and AI categorization. Features a DM-based signup workflow and a premium React dashboard for live monitoring of saved reels categorized by topic using AI.
An advanced Talk2DB platform that converts Natural Language to SQL using a RAG (Retrieval-Augmented Generation) architecture. It features strict RBAC, modular provider interfaces, and dynamic workflow execution for seamless database interaction.
Inspired by the 1930 helpline project, I developed an automated stock analysis and prediction model and paper trading platform. Uses historical data to provide insights and forecast market trends, helping to automate investment decisions.
An advanced generative AI platform for creating and managing intelligent agents. Includes a sophisticated prompt processor that enables agents to handle real-world tasks. It's like having AI on your computer, basically executing Python code autonomously.creates self agents and pre build prompts agents with custom logic of MCP for GEMINI at that time this is just a concepts.
ZenLight is a browser extension that brings AI to your fingertips — select any text on any webpage to instantly get explanations, translations, and visual context without leaving the page. Highlight a question and ZenLight automatically finds and marks the answer on the page. Plus, a distraction-free Spotlight Mode and YouTube productivity tools help you stay focused and in the zone.
A C program built during college that visualizes CPU scheduling algorithms — FCFS, SJF, SRTN, Priority (preemptive & non-preemptive), and Round Robin — using the graphics.h library. Enter your processes and instantly see a Gantt chart along with turnaround and waiting time stats. Note: Round Robin was implemented based on college-level understanding and may not cover all edge cases perfectly.
A machine learning project built at the Rajasthan Police Hackathon to detect fake complaints on the 1930 helpline. Uses scikit-learn to analyze and classify police complaint data. Not fully completed, but a great hands-on introduction to ML concepts.
A comprehensive management system for public libraries that streamlines book lending and user registrations. Dropped idea after hackathon.
Face recognition system from fetching facebook prefetched profile photo from facebook id.
An interactive game where players guess a randomly generated number with helpful hints and scoring. Build when learning JavaScript basics.
Efficient task management application with persistent storage, categories, and priority levels. Build when learning Web Development basics.
