I'm a student at the University of Minnesota majoring in IT Infrastructure. I like learning new technologies and things in general. I hope to one day inspire others, just like how others have inspired me. I'm currently looking for internships or job opportunities so feel free to reach out to me!
I wanted to practice frontend and backend development so I tried making a simple full-stack web application. It was built using Next.js (TypeScript), Tailwind CSS, Express.js, and MongoDB.
A project I made following a YouTube tutorial on Backend Development. It taught me things such as building RESTful APIs, JWT Authentication, Security, Database Modeling with MongoDB, etc.
A simple Google Extension that stimulates night shift mode by reducing blue light on any website that you visit. This helped me get into Google Chrome Extension development.
A simple portfolio website built using Next.js and Tailwind CSS. Initially, I was going to make it using React but I ended up making it using Next.js because I liked how easy the routing was.
I had the idea from learning about Embedding and I want to compare an applicant's resume against a job description to determine how well they match, using natural language embeddings and semantic similarity scoring. It uses a Sentence Transform from HuggingFace to generate the semantic embeddings and compute a cosine similarity score that reflects how well the resume fits the job. I visualized the semantic relationship between the two using a scatter plot. I used Jupyter Notebook because some of the libraries wasn't working for me locally.
QuickAI is a full-stack AI-powered SaaS platform built with the PERN stack (PostgreSQL, Express.js, React, Node.js). The platform provides a unified dashboard for multiple AI-powered tools including content generation, image manipulation, and resume analysis.
A multi-agent AI system that I built to learn about agentic AI and how specialized agents can work together to solve complex tasks. The project uses Phidata to orchestrate multiple AI agents - one for web search and another for financial data analysis. It taught me about agent coordination, tool integration with YFinance and DuckDuckGo, and how to structure AI systems that can delegate tasks intelligently. I used Groq's LLaMA 3.3 model for the AI inference.
A conversational AI assistant that reads and answers questions about PDF documents using RAG. It uses PDF knowledge base, semantic search, and vector embeddings to retrieve relevant information from documents. This project helped me learn how to combine document processing, vector databases, and conversational memory into one intelligent assistant.
A multimodal AI agent that analyzes video content and answers questions using Google's Gemini 2.0 Flash model. The agent can process uploaded videos and provide detailed insights by combining video understanding with web search capabilities through DuckDuckGo. I built it using Streamlit and Phidata's agent framework, learning how to handle video file processing, implement multimodal AI interactions, and integrate multiple tools into a single intelligent agent. I tested it with a 30-second weather news clip and it successfully extracted specific information like temperatures, locations, and forecasts from the video content.
A simplified version of Git built in Python to understand the core concepts of version control systems. It is a work in progress but I was really curious about how Git works under the hood so I decided to build my own version following this tutorial: https://www.leshenko.net/p/ugit/
An AI-powered Chrome extension that analyzes news articles for political bias, emotional framing, and fact vs. opinion distinctions in real-time. Built with Node.js, PostgreSQL, and Groq AI (Llama 3.3 70B), featuring smart caching that reduces API costs by 90%. Successfully tested on major news outlets with a focus on promoting media literacy. Built with the help of Claude Sonnet 4.5 :D
Feel free to connect with me on LinkedIn or shoot me an email!