I design and build intelligent systems — from ML/DL models to text and voice-based AI agents, and fine-tuned LLMs. With 2+ years of hands-on experience, I love turning ideas into smart, scalable solutions that push the limits of what AI can do.
Usman Haider
TrainMyAI
TrainMyAI is a user-friendly tool designed to enable anyone to train deep learning models with just a single click. Built with a simple and interactive interface powered by Streamlit, this project aims to simplify the process of model training, evaluation, and inference.
Chatical
I led a project where we built a Chrome extension that automatically captures the content of the current webpage when you open it. The content is then stored in a Pinecone vector database, allowing users to chat and ask questions about the article they're viewing. The backend was developed using FastAPI and used a Retrieval-Augmented Generation (RAG) approach for answering questions.
Student Medical Chatbot
Developed a chatbot app for MBBS students to assist with their medical books using Llama Index. Improved accuracy by fine-tuning the model and storing embeddings in OpenSearch on AWS Cloud. Built the backend in Django with proper authentication and session management.
Inventory Analysis
Led the project and trained an XGBoost model on a dataset with over 5 million rows. Developed Django APIs to streamline model training and prediction for both sales forecasting and replenishment optimization. Integrated with AWS for data storage and retrieval.
Football Analysis
Developed a football analysis system that identifies player positions and tracks their movement speed in real time during live matches. We used YOLO for object detection, OpenCV for video processing, and PyTorch for building the underlying models.
LLM Applications
This repository is dedicated to building AI agents and Retrieval-Augmented Generation (RAG) projects powered by large language models (LLMs). Our goal is to create intelligent, automated solutions for various applications.
Finetune LLM Models
In this repository, I fine-tune various large language models (LLMs), such as BERT, for tasks including sentiment analysis, question answering, summarization, text generation, text-to-text, text-to-speech, text-to-audio, audio classification, image classification, and more.
Mobile Price Prediction
The aim of this project is to predict the prices of Samsung mobile phones based on various features using two different machine learning algorithms: Linear Regression and Random Forest. The project involves data analysis, feature engineering, model building, and evaluation.
Agrefine
Agrefine is an Android application designed to assist farmers and agriculture enthusiasts in optimizing their crop cultivation and trading processes. The application is divided into three primary functionalities: Crop Prediction, Fertilizer Suggestion, Agricultural Product Marketplace
Olympics Data Analysis
The Olympic Data Analysis Web Application is a Python-based project built using Streamlit that offers comprehensive insights into Olympic data. Covering data from various Olympic events up until 2017, this web application enables users to perform in-depth analyses and explore historical trends related to the Olympics.
Usman Haider
