Portfolio

End-to-end guide to install NVIDIA drivers, CUDA 12.8, and PyTorch on Ubuntu (22.04/24.04), with ready-to-run scripts and Jupyter benchmarks to validate GPU performance (RTX 50-series tested).

AI-powered FPL Scout using historical and live data with ML models to predict player performance and optimize team selection.

RAGSearch

gpt2-from-scratch using torch

A minimal AI-powered inventory management chatbot that answers business questions and returns the exact SQL query for each answer.

Real-world MCP server / client with AI Agent capabilities using LangChain

retail-demand-forecasting desc

trasformer with torch

A web application using Retrieval-Augmented Generation (RAG) to analyze and compare survey datasets. Built with FastAPI, Nginx, and OpenAI’s models. Fully containerized with Docker for easy deploym…

AI model for voice and text interactions using FastAPI.

ICR - Identifying Age-Related Conditions Solutions

This is the legendary Titanic ML competition solution – the best, first challenge for you to dive into ML competitions and familiarize yourself with how the Kaggle platform works.

Silver medal solution + baseline of Kaggle competition Mechanisms of Action (MoA) Prediction.

Distinguish images of dogs from cats using custom CNN network and Transfer Learning with Keras.

Implement deep learning (CNN Network) and machine learning (Random Forest, KNN and Naive Bayes) to recognize digits from (0~9)

Predict whether the presence of heart disease in the patient using several machine learning algorithms with EDA && Visualizations.

21st Place Solution - Representation Is All You Need

Image similarity estimation using a Siamese Network with a triplet loss

Implement ViT Segmentation PyTorch for HuBMAP kaggle competition.

Machine Learning using NumPy

Generate Prime Numbers from 1 to 1000 from scratch in Python and SQL.

UDACITY - Advanced-Data-Analysis Track Project