Nitu Dudi | Homepage

Early Detection and Diagnosis of Alzheimer’s Disease Python | PyTorch | CNN | EfficientNetB3 | Transfer Learning

Built a deep-learning based medical image classification system using CNNs and EfficientNetB3 (ImageNet-pretrained) to detect multiple stages of Alzheimer’s disease. The model was trained using transfer learning and fine-tuned to improve classification accuracy from 94% to 97%. The pipeline includes medical image preprocessing, data augmentation, model evaluation, and performance validation using quantitative metrics to support reliable early-stage diagnosis.

Health Insurance Cost Predictor Pandas | NumPy | Scikit-learn | XGBoost | Streamlit

Developed a machine learning model on 50,000+ insurance records using Linear Regression, Random Forest, and XGBoost to predict premium costs with high accuracy. Achieved an R² score of 0.988 through extensive data cleaning, feature engineering, and hyperparameter tuning. Deployed a real-time Streamlit web application that enables users to instantly estimate premiums, significantly reducing extreme prediction errors for young customers through age-segmented retraining.

Credit Risk Evaluator Python | Pandas | Scikit-learn | Streamlit

Built an end-to-end credit risk prediction system on 50,000+ loan records, performing data cleaning, exploratory analysis, and feature engineering including LTI ratio and average DPD. Achieved AUC = 0.98 and KS > 85% using SMOTE-Tomek resampling and Optuna-tuned Logistic Regression. Deployed a Streamlit app that provides real-time credit scores (300–900) and borrower risk classification.

E-Commerce Chat Assistant Python | GenAI | ChromaDB | Streamlit

Developed an AI-powered e-commerce chatbot capable of handling small talk, FAQs, and SQL-based product queries with 95%+ routing accuracy. Integrated LLaMA-3.3 via GROQ for fast, context-aware responses and ChromaDB-powered Retrieval-Augmented Generation (RAG) for high-precision FAQ retrieval. Enabled dynamic product search over 1,000+ items through LLM-generated SQL queries.

Your-Resume (Resume Maker) MongoDB | ExpressJS | ReactJS | NodeJS

Built a full-stack resume-building platform that allows users to enter personal details and generate customized resumes using multiple templates. Implemented complete CRUD functionality to manage resume data efficiently and added a print utility to export resumes in a professional format, improving overall user experience and flexibility.

Spotify Clone HTML | CSS | Bootstrap

Designed a responsive static UI for a Spotify-like music player featuring sidebar navigation, featured playlists, and a fixed bottom playback bar. The interface closely replicates the layout and user experience of a real-world music streaming platform using clean and modern front-end design principles.

Simon Game HTML | CSS | JavaScript

Developed an interactive memory-based game that generates dynamic sound and light sequences which users must replicate to progress through increasing difficulty levels. Implemented game logic, event handling, and animations using JavaScript to create an engaging and responsive user experience.

I'm Nitu Dudi