Ghanashyam Panda

Data Aspirant

[email protected]

Resume

Hello, I am

Ghanashyam Panda.

1

Python
Project

2

Machine Learning
Projects

1

Deep Learning
Project

About Me

I'm an aspiring Data Scientist driven by a passion for transforming raw data into meaningful insights. Pursuing my B.Tech in Computer Engineering at IIIT Bhubaneswar, I discovered a natural flair for data science where logic meets real-world impact.
Skilled in Python, machine learning, deep learning, and data analysis, I’ve built several end-to-end projects that tackle real business problems, focusing on both technical accuracy and practical relevance.
Though I haven’t worked in a formal role yet, my hands-on experience has equipped me to handle the full data science pipeline from data cleaning to model deployment with a problem-solving mindset.
I'm eager to bring my energy and skills to a team where data drives decisions. Let’s connect if you're looking for someone who learns fast and delivers with purpose.

Key Skills

python

FastAPI

Streamlit

Neural network

Django

Machine Learning

Deep Learning

Image Processing

HTML

CSS

JAVA

Javascript

Jupyter Notebook

GitHub

GIT Bash

Data Analysis

My Projects

Credit Risk Model

Domain/Function: FinTech (Banking & Financial Services)

Health Insurance Premium Predictor

Domain/Function: Healthcare Analytics

Car Damage Detection Model

Domain/Function: Healthcare

Expense Tracker plus

Domain/Function: Expense Management

My Experience

Data Science Intern: EXPOSYS DATA LABS, Bengaluru

Aug 2025 - Sep 2025

Developed a robust XGBoost classification model to predict the onset of diabetes, engineering an end-to-end machine learning pipeline from data preprocessing to final model evaluation and serialization.
Engineered the final model to achieve an outstanding AUC of ~0.98 and a Gini coefficient of 0.96, indicating exceptional class separation and predictive power.
Successfully balanced the precision-recall trade-off, achieving a high precision of ~0.90 with a strong recall of 0.71 and an overall accuracy of 0.97. This was a critical improvement over alternative models that suffered from extremely low precision (~0.34) when recall was high.
Addressed significant class imbalance in the dataset using the SMOTETomek resampling technique, creating a more robust training dataset.
Systematically optimized model performance by conducting extensive hyperparameter tuning over 50 trials with Optuna and 100 iterations with RandomizedSearchCV to maximize the F1-score.
Serialized the final model, scaler, and feature list into a joblib artifact, preparing it for deployment.

Awards & Certificate

Python: Beginner to Advanced For Data Professionals

SQL for Data Science

Math and Statistics For AI, Data Science

Master Machine Learning for Data Science & AI: Beginner to Advanced

Machine Learning Workshop by AWS

Let's Connect

Feel free to get in touch with me. I am always open to discussing new projects, creative ideas or opportunities to be part of your visions.

Mail Me

[email protected]

https://www.linkedin.com/in/ghanashyam-panda-57a6aa255/

GitHub

https://github.com/ghanashyam9348

Machine Learning Project

Credit Risk Model

Domain/Function: FinTech (Banking & Financial Services)

Credit-Risk-Model is a Python-based credit risk prediction pipeline. It features data cleaning, EDA, feature engineering, SMOTE-Tomek oversampling, model training with XGBoost (plus Logistic Regression and Random Forest), hyperparameter tuning via Optuna, and a Streamlit app for real-time

View Project
on GitHUB

View Project
on Streamlit

Other Tools

Python

Jupyter Notebook

pandas

numpy

matplotlib

seaborn

scikit-learn

XGBoost

imbalanced-learn (SMOTE-Tomek)

Optuna

Streamlit

joblib

Git

GitHub

Streamlit UI of CrediSure app capturing borrower details and credit behavior to calculate loan default risk using ML-based credit scoring.

The app predicts a 65.68% loan default risk, shows a credit score of 505, and rates the applicant as “Average” based on entered financial metrics.

Project Details

Credit Risk Model is a Python-based machine learning pipeline designed to predict borrower default risk using demographic, loan and credit bureau data.
It begins with data cleaning—imputing missing values, Winsorizing outliers, and scaling features—followed by exploratory analysis with visual charts to uncover key patterns.
Feature engineering generates metrics such as debt-to-income ratios and target-encoded categories. Class imbalance is addressed with SMOTE-Tomek sampling.
Models including Logistic Regression, Random Forest and XGBoost are trained and hyperparameter-tuned via Optuna, then evaluated on recall, precision, F1-score and AUC. The chosen model is serialized with joblib and deployed in a Streamlit app that accepts user inputs and displays real-time risk scores with SHAP explanations.
The code leverages pandas, scikit-learn, imbalanced-learn, XGBoost, Optuna, joblib and Streamlit, with version control via Git.

Machine Learning Project

Health Insurance Premium Predictor

Domain/Function: Healthcare Analytics

Insurance-Premium-Predictor-App is a Python-based ML pipeline that cleans and analyzes insurance data, trains regression models with scikit-learn, and deploys the best model via Streamlit for real-time premium estimates using inputs like age, BMI, and smoker status.

View Project
on GitHUB

View Project
on Streamlit

Other Tools

Python

Jupyter Notebook

pandas

numpy

matplotlib

seaborn

scikit-learn

joblib

Streamlit

Git

GitHub

Streamlit UI for predicting health insurance premiums using user inputs like age, income, BMI, habits, and medical history.

Displays predicted health insurance premium instantly after user submits personal and medical info via Streamlit form.

Project Details

Insurance Premium Predictor App is a Python-based machine learning pipeline delivering real-time health insurance cost estimates through an interactive web interface.
It loads raw customer and charge data, cleans and transforms it by imputing missing values, handling outliers, encoding categorical fields, and scaling numerical features.
Exploratory analysis with pandas, matplotlib, and seaborn uncovers correlations that guide feature selection.
The data is split into training and test sets to train linear regression, random forest, and gradient boosting models with grid search cross-validation for hyperparameter tuning.
Models are evaluated using mean absolute error, mean squared error, and R-squared. The best model is serialized with joblib. A Streamlit app provides a form for user inputs to deliver instant premium predictions.
Dependencies are managed via requirements.txt, and version control is maintained with Git and GitHub, demonstrating a full end-to-end workflow .

Deep Learning Project

Car Damage Detection Model

Domain/Function: Healthcare

An AI web app to detect vehicle damage. 🚗 Built with PyTorch and Streamlit, it uses a deep learning model to classify damage like 'crushed' or 'breakage' from an uploaded image, providing an instant prediction.

View Project
on GitHUB

the screenshot demonstrates the application's complete workflow: a user provides an image of a damaged car, and the AI model successfully analyzes it

successfully detected the front breakage.

Project Details

This end-to-end project classifies vehicle damage from images using PyTorch for training and Streamlit for the web interface.

The training process, contained in a Jupyter Notebook (damage_prediction.ipynb), evaluates various models like custom CNNs, ResNet, and EfficientNet. These models learn to classify damage by type and location (e.g., Front_Crushed, Rear_Normal).

The best-performing model is saved as selected_model.pth. This file is then loaded by the Streamlit web app (app.py), which provides a simple UI for users to upload a vehicle image and receive an instant damage prediction.

To run the app, install the dependencies from requirements.txt and then execute streamlit run app.py in your terminal.

Python Project

Expense Tracker plus

Domain/Function: Expense Management

ExpenseTracker-Plus is a full-stack web app using FastAPI, Streamlit, and MySQL. Log, edit, and categorize income/expenses; filter by date, category, or amount; visualize interactive spending charts and generate monthly reports for real-time financial insights.

View Project
on GitHUB

Other Tools

Python 3

FastAPI

Uvicorn

Streamlit

MySQL

mysql-connector-python

Pytest

Git & GitHub

A responsive expense tracking app built with Streamlit, allowing users to log daily spending with categories, notes, and date-based summaries.

Visualizes monthly spending with bar charts and summaries, helping users track expenses clearly over time in an interactive Streamlit app.

Project Details

An end-to-end, full-stack web application that transforms expense tracking into an intuitive, data-rich experience.
Built with a FastAPI RESTful backend, a dynamic Streamlit frontend, and a MySQL database, ExpenseTracker-Plus empowers you to:
Capture & Organize entries by logging, editing, or deleting income and expense records. Assign them to default or custom categories, and add tags or notes for context.
Slice & Dice your data using versatile date ranges, categories, amounts, or custom labels to pinpoint exact transactions and trends.
Visualize spending trends with interactive bar charts for monthly totals and category breakdowns, plus generate concise monthly summaries for quick insights.
Tech Highlights: FastAPI-driven RESTful API; MySQL storage via a Python connector; responsive Streamlit interface with live updates; real-time analytics.
this project showcases full-stack Python mastery combining user-focused UI design, a robust backend, and practical finance management.