Portfolio - Kavinesh L

Kavinesh L

Aspiring AIML Engineer

Resume

2

Python Projects

2

Machine Learning Projects

1

Deep Learning Project

2

GenAI Projects

About Me

Hello! I'm Kavinesh L, an aspiring Gen AI, Data Science enthusiast with a B.E in Robotics and Automation Engineering skilled in Machine Learning, Python, and SQL. I love turning complex data into clear insights that help solve real-world problems.

Tools and Technologies:

➡️I use Python to handle data and create models that learn from it.
➡️I'm good with SQL for organizing and retrieving data.
➡️I also work with tools like Jupyter Notebooks, Pandas, and Matplotlib to analyze and show data clearly.

Areas of Expertise:

➡️Building ML, DL models that predict future trends and improve business decisions.
➡️Making data tasks faster Gen AI & Agentic AI, and more accurate with automation.
➡️Designing easy-to-understand data visualizations for better decision-making.

I believe in the power of learning from data and constantly improving. I enjoy sharing what I learn and connecting with others interested in data. Let's chat about how we can use data to make a difference!

Key Skills

Python

Machine Learning

Model Deployment

EDA

MySQL

Power BI

Data Analysis

Statistics

Deep Learning

Gen AI

Agentic AI

RAG

My Projects

E-Commerce Chatbot

Domain/Function: E-Commerce

Real Estate Assistant Tool

Domain/Function: Finance

𝐀𝐭𝐥𝐢𝐐 𝐀𝐈'𝐬 𝐂𝐚𝐫 𝐃𝐚𝐦𝐚𝐠𝐞 𝐃𝐞𝐭𝐞𝐜𝐭𝐢𝐨𝐧 𝐒𝐲𝐬𝐭𝐞𝐦

Domain/Function: Car Business

Credit Risk Machine Learning Model Project

Domain/Function: Banking and Finance

Health Insurance Cost Predictor ML Project

Domain/Function: Banking and Finance

T20-Cricket World Cup 2022 Data Analytics

Domain/Function: Sports and Entertainment

Atliq Hotels Data Analysis Project

Domain/Function: Hospitality

AtliQo Credit Card Project

Domain/Function: Banking and Finance

My Experience

AtliQ Technologies

Data Science Intern - Chennai

Nov 2025 - Dec 2025 LOE

Predictive Modeling (CodeX): Led the end-to-end development of a machine learning model for the Food & Beverage sector; performed rigorous data cleaning on 30k+ survey records using Python (Pandas) and engineered high-impact features (BSI, ZAS) to optimize model performance.
Streamlit Deployment: Built and deployed an interactive web application using Streamlit, enabling stakeholders to input demographic data and visualize real-time predictions on customer price preferences.
SQL-Based Market Analysis: Executed complex SQL queries to analyze Electric Vehicle (EV) sales trends; calculated critical metrics including CAGR, seasonal peaks, and market penetration rates to support strategic planning.
Business Impact: Translated client requirements into technical data solutions, delivering a deployed prototype and actionable insights on brand awareness and consumption patterns.

Skills: Data Science, ML, DL, Streamlit, Python and Optuna Fine Tuning.

Awards & Certificate

Math and Statistics For AI, Data Science

Python: Beginner to Advanced For Data Professionals

SQL for Data Science

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

Deep Learning: Beginner to Advanced

Natural Language Processing

Gen AI to Agentic AI with Business Projects

Gen AI & Data Science Bootcamp 3.0: With Practical Job Placement Support & Virtual Internship

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/kavineshl/

Call Me

+91 6383926186

GitHub

https://github.com/Ronin7DX

GenAI Project

E-Commerce Chatbot

Domain/Function: E-Commerce

An intelligent E-Commerce chatbot that answers any question. An AI router (LangChain & Groq) directs queries: RAG for FAQs, and Text-to-SQL for live database info on products, stock, and orders.

View Project
Engagement

View Project
on GitHUB

Other Tools

Python

Streamlit

RAG

GROQ Model

SQLite

ChromaDB

AI Router

Text-to-SQL

LangChain

AI enabled E-Commerce chatbot. An AI router uses RAG for FAQs and Text-to-SQL to query a live database for product, stock, or order info.

Streamlit chatbot for E-Commerce. An AI router (Groq) uses RAG for FAQs or Text-to-SQL to query live product data from a database.

Project Details

This project is an intelligent E-Commerce Chatbot designed to answer both general and specific customer queries.

The system's core is an AI Router (built with LangChain and Groq Llama 3.1 70B) that analyzes incoming questions and routes them to the correct backend:

RAG for FAQs: General questions (e.g., "return policy?") are handled by a Retrieval-Augmented Generation (RAG) pipeline. This system uses HuggingFace embeddings and a ChromaDB vector store to find answers in company documents.

Text-to-SQL for Live Data: Specific queries (e.g., "top 5 products?") are converted into SQL by a Groq Llama 3.1 8B model. This chain queries a live SQLite database to retrieve real-time product, stock, or order data.

This entire hybrid system is deployed in a user-friendly Streamlit web app, creating a scalable, end-to-end AI assistant.

GenAI Project

Real Estate Assistant Tool

Domain/Function: Finance

AI-powered Real Estate Assistant. Uses a RAG pipeline (LangChain, Groq, ChromaDB) to ingest live web articles and reports. Ask any question about the market and get instant, synthesized answers with sources.

View Project
Engagement

View Project
on GitHUB

Other Tools

Python

LangChain

ChromaDB

GROQ Model

Streamlit

RAG

AI Real Estate Assistant (PyCharm). Uses a RAG pipeline (LangChain, Groq, ChromaDB) to ingest web articles and answer market questions.

Streamlit app for Real Estate. Chat with live web articles. Powered by a RAG pipeline (LangChain, Groq, ChromaDB) for instant market answers.

Project Details

This AI-powered Real Estate Research Tool automates market analysis by allowing users to "chat" with live web content.

It's built on an end-to-end Retrieval-Augmented Generation (RAG) pipeline. The user inputs URLs (news, reports), and the system uses LangChain's UnstructuredURLLoader to scrape and process the text. This content is then chunked, embedded (using HuggingFace models), and stored in a ChromaDB vector database.

When a user asks a question, the RAG system retrieves relevant context from ChromaDB. This context is fed to a high-speed Groq Llama 3.1 70B model, which synthesizes the information to provide a single, accurate answer.

The entire application is deployed as an interactive Streamlit web app, demonstrating a practical solution for turning unstructured web data into actionable insights.

Deep Learning Project

𝐀𝐭𝐥𝐢𝐐 𝐀𝐈'𝐬 𝐂𝐚𝐫 𝐃𝐚𝐦𝐚𝐠𝐞 𝐃𝐞𝐭𝐞𝐜𝐭𝐢𝐨𝐧 𝐒𝐲𝐬𝐭𝐞𝐦

Domain/Function: Car Business

Car Damage Detection project classifies vehicle damage by location and type using a fine-tuned ResNet50 model, achieving ~81% accuracy. Deployed with a Streamlit UI for interactive analysis and a FastAPI API for programmatic access , it automates initial insurance assessments.

View Project
Engagement

View Project
on GitHUB

View Project
on Streamlit

Other Tools

ResNet50

Optuna

Streamlit

FastAPI

PyTorch

Jupyter Notebook

Python Programming

Generated a confusion matrix to evaluate a vehicle damage classifier, visualizing prediction accuracy across six distinct front and rear damage types.

Developed a Streamlit web app for the Car Damage Detection project, featuring a simple drag-and-drop UI for users to easily upload car images.

Project Details

The Car Damage Detection Project automates insurance assessments by classifying vehicle damage from images into six categories based on location (front/rear) and type (breakage, crushed, normal). Built in Python, it uses a fine-tuned ResNet50 model trained via transfer learning, achieving an 81% validation accuracy. The model's hyperparameters were optimized using the Optuna framework to enhance performance. For practical application, the project is deployed as an interactive Streamlit web app for user-friendly demonstrations and a scalable FastAPI REST API for programmatic access. Future enhancements include adding bounding box detection with models like YOLO and classifying damage severity as of September 20, 2025.

Project video

Machine Learning Project

Credit Risk Machine Learning Model Project

Domain/Function: Banking and Finance

Credit Risk Modeling Project predicts loan defaults using Logistic Regression, trained on 50,000 records from customer, loan, and bureau data. Achieves 97.5% AUC, 85.9% KS, and 0.96 Gini. Includes Streamlit UI for real-time credit scoring (300–900) and ratings.

View Project
Engagement

View Project
on GitHUB

View Project
on Streamlit

Other Tools

Python

EDA

EXCEL

Pandas

Numpy

Scikit-learn

Matplotlib

Seaborn

Streamlit

KDE plots of all columns showing default (0/1) distributions for age, income, loan amount, tenure, and more, with overlaid density curves for analyze.

4 / 4 Loki Finance Credit Risk Modelling UI on Streamlit, showing input fields for age, income, loan details, and a "Calculate Risk" button.

Project Details

The Credit Risk Modeling Project enhances loan approvals by predicting default probabilities and assigning credit scores (300–900) inspired by the CIBIL system. Built in Python, it uses a Logistic Regression model trained on 50,000 records each from loan, customer, and bureau data via Pandas, achieving 97.5% AUC, 85.9% KS, and 0.96 Gini. It employs NumPy, Matplotlib, and Seaborn for preprocessing and visualizations, with numerical (age, loan tenure, credit utilization) and categorical (residence type, loan purpose) inputs. Future ML Ops integration could boost scalability as of August 14, 2025.

Project video

Machine Learning Project

Health Insurance Cost Predictor ML Project

Domain/Function: Banking and Finance

This project develops an explainable ML model using segmented Linear Regression to predict health insurance premiums, deployed via Streamlit for real-time underwriting.

View Project
Engagement

View Project
on GitHUB

View Project
on Streamlit

Other Tools

Python

EDA

Excel

Pandas

Numpy

Matplotlib

Seaborn

Streamlit

Scikit-learn

A correlation heatmap visualization is generated to analyze the relationships between features like age, income, and BMI category, identifying potenti

A Streamlit app is shown, with a user interface for underwriters to input client data to get a real-time predicted premium.

Project Details

This project delivers a robust machine learning solution for predicting health insurance costs. It leverages Python with Pandas for data preprocessing, including a critical segmentation strategy based on age to address data error patterns. The core of the solution involves training two distinct Linear Regression models and optimizing them using hyperparameter tuning. The predictive models achieve a high level of accuracy, with 95% of predictions having an error rate below 10%. A user-friendly Streamlit application provides a real-time, interactive interface for insurance underwriters, facilitating immediate premium predictions. The models and scalers are serialized using joblib and stored as artifacts for efficient deployment. This project demonstrates proficiency in the full ML lifecycle, from data engineering and modeling to application deployment.

Project video

T20-Cricket World Cup 2022 Data Analytics

Domain/Function: Sports and Entertainment

Power BI dashboard analyzing 2022 Men's T20 Cricket World Cup player performance using data scraped from ESPNcricinfo and BrightData, with Python, Pandas, and Beautiful Soup.

View Project
Engagement

View Project
on GitHUB

View Project
on Power BI Website

Other Tools

Power BI

Excel

Power Query Editor

This Power BI dashboard from the T20 World Cup 2022 project displays batting avg (20.16), strike rate (117.08), and boundary % (50.34).

This Power BI dashboard analyzes power hitters, aiding selection of top performers like Kusal Mendis and Rilee Rossouw for the best 11.

Project Details

This project creates a Power BI dashboard to evaluate player performances in the 2022 Men's T20 Cricket World Cup. Data scraped from ESPNcricinfo and BrightData was processed using Python and Beautiful Soup in Jupyter Notebook, converted to CSV, and cleaned with Pandas. Power Query handled transformations, linking datasets via team and match IDs. DAX measures like Total Runs, Batting Average, and Economy were used for analysis. The dashboard visualizes key metrics, enabling selection of the best 11 players based on performance criteria. Tools include Jupyter Notebook, Python, Pandas, Beautiful Soup, and Power BI, managed in an Anaconda environment. The project offers a robust framework for cricket analytics and team selection insights.

Project video

Python Project

Atliq Hotels Data Analysis Project

Domain/Function: Hospitality

This project performs a data analysis on fictional Atliq Hotels booking data, providing key business insights on revenue, occupancy, and platform performance using Python, Pandas, and data visualization libraries like Matplotlib and Seaborn.

View Project
Engagement

View Project
on GitHUB

View Project
on

Other Tools

Excel

Pandas

Numpy

Seaborn

Matplotlib

Analysis of June occupancy across cities, showing unique dates, room categories, and occupancy rates for properties like Atliq Grands and Atliq Blu.

Pie chart displaying revenue realized per booking platform, including makeyourtrip, direct online, journey, and others, with clear segment labels.

Project Details

This project performs a comprehensive data analysis of booking data for Atliq Hotels, a fictional hospitality chain. It leverages Python with Pandas for data manipulation and Matplotlib/Seaborn for visualizations. The analysis integrates six distinct datasets to derive key business insights on revenue, occupancy rates, and platform performance. It calculates metrics such as occupancy percentages, revenue realized, and cancellation rates, while also addressing data quality issues. The findings, presented through visualizations like revenue distribution pie charts and occupancy trends, provide actionable intelligence for strategic decision-making, such as optimizing booking platforms and pricing. The project is an excellent demonstration of descriptive analytics skills applied to the hospitality industry.

Python Project

AtliQo Credit Card Project

Domain/Function: Banking and Finance

The credit card project at AtliQo Bank comprises two key phases: market identification and trial. This initiative aims to leverage mathematical and statistical concepts to analyze data , and spending patterns in order to identify the target audience for the credit card.

View Project
Engagement

View Project
on GitHUB

Other Tools

Excel

Statistics

EDA

MySQL

Visualized credit card campaign data, analyzing average transaction amounts by payment type and platform. This code snippet shows bar plots for EDA.

Performed a one-tailed Z-test with a 'larger' alternative hypothesis to check if the test group's average transaction was significantly higher.

Project Details

Situation:
AtliQo Bank was launching a new credit card and needed to identify the right target market and test a new campaign strategy.

Task: My goal was to use data analysis to pinpoint the ideal audience and then apply hypothesis testing to evaluate the effectiveness of a trial campaign.

Action: I first cleaned and validated the dataset, using SQL and visualizations to identify that the 18-25 age group, despite being a target, showed low credit card usage. I then designed a trial campaign, divided participants into control and target groups, and used a z-test with a sample size of 100 to statistically measure the new card's performance.

Result: The project's structured approach, which combined data analysis and hypothesis testing, enabled AtliQo Bank to make informed, data-driven decisions, resulting in a successful strategy for the new credit card launch.