Portfolio - Sarthak Maddi

Machine Learning Project

Healthy Bytes: Smart Diet Guide

Domain/Function: AI / Nutrition / Food Recommendation System

An AI-powered diet assistant that predicts whether Indian dishes are healthy, moderate, or unhealthy and recommends personalized meals based on nutritional filters like calories, sugar, and protein.

View Project
Engagement

View Project
on GitHUB

View Project
on Streamlit Cloud

Other Tools

FastAPI

Streamlit

scikit-learn

Pandas

Matplotlib

CSV tools

Homepage lets users input a food item, instantly predicts its health category: Healthy, Moderate, or Unhealthy using AI model.

Based on selected health label, users get smart recommendations of Indian dishes tailored to their health goals and nutrition needs.

Project Details

Healthy Bites is a full-stack AI-based food recommendation system built with FastAPI and Streamlit. It predicts the healthiness level of Indian dishes—Healthy, Moderate, or Unhealthy—based on nutritional content. Users can get personalized dish suggestions by applying filters for calories, protein, and sugar. The system uses a Kaggle Indian food nutrition dataset and an ML classification model with preprocessing and feature scaling. The backend serves real-time predictions and recommendations, while the frontend offers an engaging UI with health badges, filters, CSV export, and responsive layout. The project emphasizes clean architecture, modular code, and deployability with render and streamlit cloud.

Project video

Machine Learning Project

Credit Risk Modelling

Domain/Function: Fintech / Credit Risk Scoring

A Streamlit-based credit risk predictor using financial and behavioral features. Trained on anonymized lending data using logistic regression and MinMax scaling to score user risk in real-time.

View Project
Engagement

View Project
on GitHUB

View Project
on Streamlit Cloud

Other Tools

Streamlit

Scikit-Learn

Pandas

Jupyter Notebook

XGBoost

Desktop view of CredSure app with sidebar info and user input fields for credit profiling and real-time risk prediction.

Mobile-responsive interface showing form inputs for credit profiling with clean UI and compact layout.

Project Details

CredSure is an intelligent credit risk assessment application that leverages machine learning to assist financial institutions in making informed lending decisions. Built using logistic regression, the model analyzes engineered features from financial and behavioral data to classify users into risk categories. It computes the probability of default and maps it to a credit score ranging from 300 to 1000, assigning a qualitative risk grade such as “Low Risk” or “High Risk.” The system includes thorough EDA, missing value handling, feature scaling with MinMaxScaler, and model evaluation metrics like ROC AUC. The final model is deployed on a responsive Streamlit dashboard, allowing real-time scoring via an intuitive interface. The app also explains the prediction confidence and highlights the user’s risk band. Future improvements include better explainability (e.g., SHAP), and real-time validations for live deployment.

Machine Learning Project

Health Insurance Premium Prediction

Domain/Function: Insurance / Cost Prediction

A regression model that predicts health insurance premium costs using individual lifestyle and demographic data. Offers an interactive dashboard to visualize health risk factors.

View Project
Engagement

View Project
on GitHUB

View Project
on Streamlit Cloud

Other Tools

Streamlit

Matplotlib

Sklearn

pandas

Numpy

Python

Full desktop view with health, lifestyle, and region inputs, plus real-time insurance premium prediction and UI instructions.

Mobile-friendly layout of This app showing fields for age, dependents, and income to predict insurance premium.

Project Details

This is a regression-based machine learning app that predicts medical insurance premiums based on user-specific attributes such as age, BMI, gender, smoking status, number of children, and residential region. The app is built with a focus on usability, offering a clean Streamlit interface where users input personal health data and instantly receive premium estimates. Behind the scenes, the project performs detailed exploratory data analysis (EDA), handles outliers and skewed distributions, and applies hyperparameter tuning for optimal prediction accuracy. It also includes visualizations for correlation insights and model evaluation metrics such as MAE and RMSE. This App bridges the gap between user-friendly design and robust ML modeling for real-world applications in the insurance sector.

SQL Project

BlogSphere

Domain/Function: Blogging Platform

BlogSphere is a simple blogging platform where users can sign up, post and edit blogs, and view others' profiles. Each profile displays a name, caption, hobbies, and followers. Built with Node.js, EJS, and MySQL, the app supports full blog CRUD functionality.

View Project
on GitHUB

Other Tools

Node.js

EJS

MySQL

HTML

CSS

JAVASCRIPT

Shows blog creation, post editing, and login/signup UI—highlighting blog CRUD features and authentication in a minimal EJS-rendered interface.

Displays user profile with name, hobbies, and blog stats, along with clean blog feed and navigation bar for seamless access and management.

Project Details

BlogSphere is a lightweight, full-stack blogging platform developed using Node.js, EJS, and MySQL. It allows users to sign up, log in, and manage personal blogs with CRUD operations. Each profile showcases the user's name, caption, hobbies, followers, and published blogs. The app supports user authentication, dynamic routing, and clean UI rendering. Designed for simplicity and usability, BlogSphere emphasizes scalable backend logic and responsive front-end templating for a smooth blogging experience.

Simon Says Game

Domain/Function: Memory Game

A classic memory-based color sequence game built with JavaScript. Tests how long users can remember the growing pattern of colors. It is one of the starting Project Of my Career in Web Dev.

View Project
on GitHUB

Other Tools

HTML

CSS

JavaScript

It shows the Simon Says game after a wrong move. The screen displays the current score and a "Retry" button to start the game again.

This screenshot shows the game reset screen. It displays a "Reset" button to restart the game and shows the highest score achieved so far.

Project Details

Simon Says is a web-based memory game where players replicate an increasingly complex sequence of colors and sounds. Built with HTML, CSS, and JavaScript, it features engaging animations, interactive sound feedback, and a responsive layout. The game challenges users to stay focused and improve their recall skills with each level. It’s a fun and educational tool for practicing pattern recognition, logical sequencing, and concentration, while also demonstrating core front-end development skills.

Hello, I am

Sarthak Maddi

3

1

About

Key

My

Healthy Bytes: Smart Diet Guide

Credit Risk Modelling

Health Insurance Premium Prediction

BlogSphere

Simon Says Game

Awards

Contact Me

Email

Github

Linkedin

Call

Sarthak Maddi

Download Resume