Portfolio - Kaushik S

Kaushik S

AI / Machine Learning Engineer

Resume

2

Machine Learning Projects

1

Deep Learning Project

About Me

Hello! I'm Kaushik, an AI / Machine Learning Engineer with a strong foundation in programming, data analysis, and problem-solving. I enjoy working with data, understanding patterns, and building practical solutions that improve real-world processes.

Tools & Technologies:
• Python for development, data processing, and automation
• SQL for managing and querying structured data
• Experience with Pandas, NumPy, Scikit-learn, Matplotlib, and Jupyter Notebook

Areas of Focus:
• Developing end-to-end ML models and analytical workflows
• Writing clean, efficient code to solve business and technical problems
• Creating clear visualizations and reports to support decision-making

I’m open to opportunities across AI, ML, data, and software-focused roles where I can apply my technical skills and continue growing as an engineer.

Key Skills

Machine Learning

Python Programming

Data Analysis

SQL

Feature Engineering

Model Development

My Projects

Real-Time ASL Gesture Recognition (Hand-landmark based)

Domain/Function: Assistive Technology / Accessibility

Credit Risk Modelling – Loan Default Prediction

Domain/Function: Finance / Credit Risk

Health Insurance Premium Prediction – Regression Model

Domain/Function: Insurance / Healthcare Analytics

My Experience

AI – Machine Learning Engineer Intern – Rooman Technologies, Bengaluru

Sep 2024 – Mar 2025

• Worked on end-to-end ML workflows including data preprocessing, feature engineering, model development, and evaluation using Python and Scikit-learn.

• Built a customer insights analysis solution as part of IBM SkillsBuild, applying Watson AI tools to extract meaningful behavioral patterns.

• Applied ML techniques in real-world OJT scenarios, focusing on clean data pipelines, model tuning, and practical deployment considerations.

• Conducted exploratory data analysis (EDA) to uncover trends, identify anomalies, and support data-driven decision-making.

• Strengthened applied ML and engineering skills through industry-backed training under IBM, NSDC, NASSCOM, and Wadhwani Foundation programs.

Awards & Certificate

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

Cognitive Customer Insights with Watson AI

AI – Machine Learning Engineer

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.

Download Resume

Resume

Mail Me

[email protected]

https://www.linkedin.com/in/kaushiks-info/

GitHub

https://github.com/kaushiks-info

Deep Learning Project

Real-Time ASL Gesture Recognition (Hand-landmark based)

Domain/Function: Assistive Technology / Accessibility

Real-time ASL recognition using Mediapipe hand landmarks and a lightweight TensorFlow model. The system converts webcam gestures into text (and speech) with a fast, low-latency pipeline suitable for live demos.

View Project
on GitHUB

Mediapipe hand landmark extraction used for preprocessing and model training in the ASL recognition system.

Real-time ASL gesture recognition demo showing webcam input, predicted label, and hand landmark tracking.

Project Details

This project recognizes American Sign Language (ASL) gestures in real time using Mediapipe to extract 21 hand landmarks and a custom TensorFlow model trained on landmark data. The workflow includes landmark data collection, preprocessing, training, and an inference script for live webcam prediction. A lightweight approach (numeric landmark arrays instead of raw images) keeps the model small and performant. The demo shows text output and optional speech synthesis for accessibility. The codebase includes scripts for data collection, training, and running the recognizer, making it easy to retrain or extend to new gestures.

Machine Learning Project

Credit Risk Modelling – Loan Default Prediction

Domain/Function: Finance / Credit Risk

Credit risk model that predicts the probability of loan default using customer financial and behavioral data. Includes data cleaning, EDA, feature engineering, and classification modelling to support lending decisions.

View Project
on GitHUB

View Project
on Web App

EDA visualization showing age distribution patterns for defaulters vs non-defaulters.

ROC curve showing model performance with an AUC score of 0.98.

Project Details

This project focuses on predicting loan default risk using machine learning classification techniques. The workflow includes data preprocessing, handling missing values, encoding categorical variables, and performing exploratory data analysis to identify key factors influencing loan repayment behavior. Feature engineering techniques such as scaling, outlier detection, and correlation analysis were applied to improve model quality. Multiple algorithms like Logistic Regression, Random Forest, and XGBoost were evaluated using accuracy, precision, recall, ROC-AUC, and confusion matrix. The final model helps financial institutions assess borrower risk and make informed lending decisions.

Machine Learning Project

Health Insurance Premium Prediction – Regression Model

Domain/Function: Insurance / Healthcare Analytics

A regression-based machine learning model that predicts health insurance premiums using customer demographics and lifestyle-related features. Includes data cleaning, EDA, feature engineering, and model evaluation.

View Project
on GitHUB

View Project
on Web App

Correlation heatmap showing relationships between customer features and the annual insurance premium.

Feature importance chart from the XGBoost model highlighting key drivers of health insurance premium prediction.

Project Details

This project focuses on predicting health insurance premiums using regression algorithms. The dataset includes customer demographics, lifestyle attributes, medical history, and other risk factors. The workflow includes data preprocessing, handling missing values, outlier treatment, and exploratory data analysis to understand key drivers affecting premium amount. Multiple models such as Linear Regression, Random Forest Regressor, and XGBoost were trained and compared using RMSE, R², and MAE. Feature engineering techniques and hyperparameter tuning were applied to improve performance. The final model helps insurers estimate premiums more accurately based on applicant profiles.