Vaibhav Garg

Aspiring Data Scientist

[email protected]

Resume

Hello, I am

Vaibhav Garg.

1

Python
Project

3

Machine Learning
Projects

2

Deep Learning
Projects

5

GenAI
Projects

About Me

👋 Welcome to my portfolio!

🧠 I’m a Computer Science graduate specializing in AI/ML, passionate about building intelligent systems that solve real-world problems through clean code, data, and automation.⚙️

My journey began with a strong foundation in Python, SQL, and core libraries like Pandas, Scikit-learn, and PyTorch.📊 Over time, I’ve expanded into Deep Learning, Natural Language Processing, and Generative AI, crafting end-to-end solutions—from dashboards to LLM-powered applications—using tools like Streamlit and ChromaDB.🌐

My projects reflect hands-on experience across core AI domains—ranging from classical Machine Learning and Deep Learning to cutting-edge Generative AI tools like LLMs 🤖,embeddings, vector databases, and conversational interfaces 💬—bridging theory

🚀 I’m always learning, experimenting with modern ML and GenAI tools, and turning ideas into useful, data-driven solutions.💡 Let’s connect or explore my work to see what I’ve been building.🌟

Key Skills

Python

MySQL

Exploratory Data Analysis (EDA)

Data Visualization

Statistical Data Analysis

Machine Learning

Deep Learning

Natural Language Processing

Generative AI

Agentic AI

My Projects

Expense Tracking System

Domain/Function: Personal Finance Management

Health Insurance Premium Predictor

Domain/Function: Healthcare

Credit Risk Evaluator

Domain/Function: NBFC Credit Risk Modelling

Beverage Price Prediction

Domain/Function: FMCG Analytics

Car Damage Detector

Domain/Function: Insurance Technology

Fruit Freshness Classifier

Domain/Function: Food Technology

Real Estate Research Tool

Domain/Function: Real Estate Analytics

E-commerce Chat Assistant

Domain/Function: E-commerce

Healthcare QA Assistant

Domain/Function: Healthcare

Human Resource Assistant Tool

Domain/Function: Human Resource Technology

AI Knowledge Explorer

Domain/Function: Conversational AI & Information Retrieval

My Experience

Data Science Intern - Atliq, Remote

Jul 2025 - Aug 2024

● Conducted Nova Mart promotion campaign analysis, involving data cleaning and exploratory data visualization using Pandas and Matplotlib, to extract actionable insights for sales and marketing decisions.

● Built a Beverage Price Prediction model leveraging XGBoost for regression tasks, and integrated MLflow for experiment tracking, model management, and reproducibility.

● Developed a Fruit Freshness Classifier using Convolutional Neural Networks (CNNs) and applied Transfer Learning with ResNet50, achieving high accuracy in image-based freshness detection.

● Designed a Healthcare QA Assistant powered by Generative AI, integrating PubMed as a knowledge source and ChromaDB for retrieval-augmented generation (RAG) to deliver accurate, con medical responses.

AI/ML Intern - IBM SkillsBuild, Remote

Jun 2024 - Jul 2024

● Contributed to the development of AquaGuardianBot, an AI-powered chatbot using IBM Watson Studio and NLP, reaching 500+ users with personalized water conservation guidance.

● Integrated real-time regional water usage and quality data, improving response relevance by 30% and enabling localized, actionable insights.

● Implemented multilingual and cross-platform support (5+ languages), boosting accessibility and increasing user engagement by over 60%.

● Enhanced overall user experience and adoption by combining AI-driven personalization, real-time insights, and language inclusivity.

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

Deep Learning: Beginner to Advanced

Natural Language Processing

Gen AI to Agentic AI with Business Projects

Gen AI & Data Science Bootcamp 2.0: With Practical Job Placement Support & Virtual Internship

Data Science Internship 1 at Atliq Technologies

Data Science Internship 2 at Atliq Technologies

AI/ML Internship at IBM SkillsBuild

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/vaibhavgarg-cse/

GitHub

https://github.com/vaibhavgarg2004

Python Project

Expense Tracking System

Domain/Function: Personal Finance Management

Built a personal finance dashboard using Streamlit and FastAPI to track, visualize, and analyze expenses, reducing manual effort by 40% compared to Excel workflows.

View Project
Engagement

View Project
on GitHUB

Other Tools

SQL

Streamlit

FastAPI

Add/Update Expenses Tab

Analytics By Category Tab

Project Details

Managing expenses manually can be time-consuming and error-prone. To solve this, I built a full-stack Expense Tracking System that lets users log, update, and analyze expenses through an interactive dashboard.

The system uses Streamlit for the frontend, FastAPI for the backend, and MySQL for persistent storage. It supports local deployment within minutes and real-time updates.

Key Features:
• Add or update expenses by date and category
• View monthly and category-wise summaries
• Visualize trends with dynamic charts
• Reduce manual effort by ~40%

This project demonstrates backend–frontend integration, SQL-based querying, and deployable Python development. It fits within the Personal Finance Management domain and serves as a base for more advanced budgeting tools.

Machine Learning Project

Health Insurance Premium Predictor

Domain/Function: Healthcare

Developed a Streamlit-based web app to predict health insurance premiums using machine learning. Included end-to-end preprocessing, feature engineering, and real-time prediction capabilities.

View Project
Engagement

View Project
on GitHUB

View Project
on Streamlit

Other Tools

Python

Scikit-learn

Streamlit

XGBoost

Prediction for Young User

Prediction for Adult User

Project Details

Estimating insurance premiums manually can be inaccurate and time-consuming. To solve this, I built a Health Insurance Premium Prediction system using machine learning and deployed it as a responsive Streamlit app.

The project features a complete ML pipeline with Pandas, Scikit-learn, and custom feature engineering. It supports real-time user input and delivers instant premium estimates based on demographic and health data.

Key Features:
• 10+ preprocessing steps including VIF-based reduction and outlier removal
• Reduced prediction error for young users from 73% to 2% via age-based segmentation
• Achieved R² of 0.988 on test data
• Real-time output in under 1 second

This project highlights applied ML in the insurance tech domain, combining data preprocessing, model optimization, and frontend deployment.

Machine Learning Project

Credit Risk Evaluator

Domain/Function: NBFC Credit Risk Modelling

Developed a credit risk prediction system using machine learning to assess loan default probability, offering real-time credit scoring and risk categorization through a responsive Streamlit web app.

View Project
Engagement

View Project
on GitHUB

View Project
on Streamlit

Other Tools

Python

Scikit-learn

XGBoost

Streamlit

Low-risk applicant with strong profile

High-risk applicant with weak history

Project Details

Assessing loan risk manually is time-consuming and inconsistent. To address this, I developed a machine learning–based Credit Risk Evaluator that predicts the likelihood of loan default using borrower data and financial history.

The system is built with Streamlit for the frontend and integrates a full preprocessing and modeling pipeline in Scikit-learn. It uses engineered features like LTI Ratio and Average DPD to improve recall on risky applicants and supports real-time credit scoring.

Key Features:
• Predicts loan default risk using financial data
• Risk categorization from 300–900 based on model output
• Improved recall on default class with SMOTE Tomek + Optuna tuning
• Real-time web app for scoring new applicants

This project falls under the Credit Risk Modeling and Financial Risk Analytics domains and demonstrates the use of ML for practical, high-impact decision-making in finance.

Machine Learning Project

Beverage Price Prediction

Domain/Function: FMCG Analytics

Developed a Streamlit-based web app to predict beverage price ranges using machine learning. Implemented end-to-end preprocessing, feature engineering, and real-time prediction based on customer demographics, preferences, and consumption behavior.

View Project
Engagement

View Project
on GitHUB

View Project
on Streamlit

Other Tools

Python

Scikit-learn

Streamlit

XGBoost

High-income urban professional with low consumption frequency

Lower-income rural professional with low consumption frequency

Project Details

Assessing beverage pricing manually is time-consuming and often inaccurate. I developed a machine learning–based Beverage Price Prediction system that estimates optimal market price ranges from customer demographics, preferences, and lifestyle choices.

Built with Streamlit, it integrates full preprocessing and modeling pipelines in Scikit-learn. Engineered features like cf_ab_score, zas_score, and bsi improve prediction accuracy across consumer segments and support real-time estimation.

Key Features:
• Predicts price ranges (₹50–100, ₹100–150, ₹150–200, ₹200–250) from user inputs
• Handles demographic, behavioral, and preference data with one-hot encoding
• Compares multiple models and tracks performance with MLflow
• Interactive Streamlit web app for instant predictions

This project falls under FMCG Analytics and Predictive Retail Pricing, demonstrating ML for practical, data-driven decision-making in the beverage industry.

Deep Learning Project

Car Damage Detector

Domain/Function: Insurance Technology

Built a Streamlit web app for automated car damage detection using deep learning. Included end-to-end image preprocessing, seamless user uploads, and real-time classification into six damage categories for instant assessments.

View Project
Engagement

View Project
on GitHUB

View Project
on Streamlit

Other Tools

Python

PyTorch

Streamlit

ResNet50

Classification output of car image with Noticeable deformation at the front.

Classification output of car image with Undamaged rear section.

Project Details

Assessing car damage manually can be inconsistent and slow. To address this, I developed an automated Car Damage Detection system using deep learning, deployed through an interactive Streamlit app.

The project includes a complete computer vision pipeline with PyTorch and torchvision, applying extensive data augmentation and leveraging ResNet50 for image classification. Users can upload car images and receive instant damage category predictions.

Key Features:
• Applied 4+ image augmentation techniques to boost model robustness
• Improved classification accuracy from 57.7% to 79.5% with regularization and Optuna-based hyperparameter tuning
• Supports real-time inference directly in the web app
• Predicts one of six damage categories (e.g., Front Crushed, Rear Normal)

This project demonstrates practical deep learning for the automotive and insurance sectors, combining advanced modeling, optimization, and a user-friendly deployment.

Deep Learning Project

Fruit Freshness Classifier

Domain/Function: Food Technology

Developed an interactive Streamlit application that leverages deep learning to evaluate fruit freshness from images. Integrated automated preprocessing, intuitive image uploads, and rapid classification into freshness levels, enabling quick and consistent quality checks.

View Project
Engagement

View Project
on GitHUB

View Project
on Streamlit

Other Tools

Python

PyTorch

Streamlit

ResNet50

Classification output of fruit image indicating freshness with vibrant appearance.

Classification output of fruit image indicating staleness with visible deterioration.

Project Details

Classifying fruit quality manually can be subjective and slow. To address this, I developed an automated Fruit Classification system using deep learning, deployed via an interactive Streamlit app.

The project features a complete computer vision pipeline with PyTorch and torchvision, using multiple augmentation techniques and ResNet50 for classification. Users can upload fruit images and instantly receive freshness category predictions.

Key Features:
• Applied 4+ augmentations to improve model robustness
• Achieved >95% accuracy in under 3 epochs
• Supports real-time inference in the web app
• Predicts one of several fruit categories (e.g., Fresh Banana, Spoiled Mango)

This project showcases practical deep learning for agriculture and food quality control, combining advanced modeling with a user-friendly deployment.

GenAI Project

Real Estate Research Tool

Domain/Function: Real Estate Analytics

Designed a Streamlit-based AI tool for real estate research using RAG to extract insights from news articles. Integrated LLaMA3 via GROQ, MiniLM embeddings, and ChromaDB for semantic search over scraped content.

View Project
Engagement

View Project
on GitHUB

View Project
on Streamlit

Other Tools

Python

ChromaDB

LangChain

Groq

Streamlit

Displays a user querying mortgage interest rates. The tool processes multiple URLs and returns an accurate, source-linked response using RAG.

Shows a user asking about luxury home sales trends. The app extracts key insights from real estate news and presents a concise answer with attribution

Project Details

Manually analyzing real estate news for market trends can be time-consuming and inconsistent. To streamline this, I built a Real Estate Research Tool using a conversational AI interface powered by LLaMA-3, deployed through an interactive Streamlit app.

The tool leverages Retrieval-Augmented Generation (RAG), semantic search, and live web scraping to provide grounded, explainable answers based on real estate news articles—making research faster and more reliable.

Key Features:
• Scrapes article content from user-provided URLs using BeautifulSoup
• Uses LangChain and MiniLM to chunk and embed text for semantic retrieval
• Stores and queries embeddings in ChromaDB for efficient RAG-based answering
• Displays generated answers along with source links in a clean Streamlit UI

This project demonstrates practical GenAI use in real estate analytics, integrating LLM reasoning, vector databases, and frontend deployment.

GenAI Project

E-commerce Chat Assistant

Domain/Function: E-commerce

Developed a Streamlit-based conversational AI app for e-commerce platforms to assist with FAQs, product discovery, and small talk. Integrated LLaMA3 via GROQ, semantic routing, and live SQL queries on a product database.

View Project
Engagement

View Project
on GitHUB

View Project
on Streamlit

Other Tools

Python

SQL

ChromaDB

Groq

Streamlit

Shows the chat assistant handling Small Talk, FAQs, and Product Search—delivering smart, contextual responses for a seamless shopping experience.

Project Details

Manually handling customer queries and product search on e-commerce platforms can be inefficient and limited. To solve this, I built an E-commerce Chat Assistant using a conversational AI interface powered by LLaMA-3 and deployed it via a fast, responsive Streamlit app.

The project combines Retrieval-Augmented Generation (RAG), SQL querying, and semantic intent routing to support dynamic assistance for FAQs, product discovery, and casual conversation—all without a backend server.

Key Features:
• Semantic intent routing to classify queries into FAQ, SQL, or small talk
• Dynamic product filtering via SQL on a local SQLite database
• RAG pipeline using ChromaDB and Sentence Transformers for policy-related questions
• Deployed using Streamlit with interactive product cards and dynamic outputs

This project demonstrates practical GenAI integration in e-commerce, showcasing full-stack capabilities including LLM-based logic, semantic search, and frontend deployment.

GenAI Project

Healthcare QA Assistant

Domain/Function: Healthcare

Developed a Streamlit-based healthcare QA assistant that retrieves PubMed articles, embeds them into ChromaDB, and answers medical queries using LLaMA3 via GROQ. Integrated semantic search, RAG pipeline, and transparent citation of article metadata.

View Project
Engagement

View Project
on GitHUB

View Project
on Streamlit

Other Tools

Python

ChromaDB

Groq

Streamlit

User asks about vitamin D and depression; the app provides an evidence-based summary with cited studies.

User asks about aerobic exercise and blood pressure; the app gives a research-backed answer with PubMed citations.

Project Details

Sometimes, it can be difficult for doctors and medical researchers to provide deterministic answers during patient consultations or clinical discussions. To address this, I developed a Healthcare QA Assistant that retrieves evidence-based information directly from PubMed and presents concise, cited answers.

The system combines automated literature retrieval, semantic search, and RAG (Retrieval-Augmented Generation) to support clinicians with reliable medical insights.

Key Features:
• Automated PubMed retrieval of abstracts and metadata for medical queries
• Semantic chunking and embeddings using Sentence Transformers + ChromaDB
• RAG pipeline powered by LLaMA3 (via GROQ) for natural language QA
• Transparent answers with citations including authors, journals, and publication dates

This project demonstrates the application of GenAI in healthcare research, showcasing skills in LLM integration, semantic search, and domain-specific QA system design.

GenAI Project

Human Resource Assistant Tool

Domain/Function: Human Resource Technology

Developed an agentic HR assistant using Claude MCP to automate onboarding, leave, meetings, and ticketing. Built a FastMCP server in Python with modular tools, Pydantic validation, secure email and dummy data generation for easy demo.

View Project
Engagement

View Project
on GitHUB

Other Tools

Python

FastMCP

Claude Desktop

smtplib

Employee Onboarding Workflow in Claude

Leave Request and History Process in Claude

Project Details

Manually managing HR tasks like onboarding, leave tracking, and ticketing is repetitive and error-prone. To streamline this, I built an Agentic HR Assistant using Claude’s MCP protocol, designed to automate core workflows through natural language interaction.

The assistant integrates with Claude Desktop and responds to user prompts by invoking backend tools that handle employee operations—powered by a FastMCP server and modular service managers.

Key Features:
• Automates onboarding, meeting scheduling, leave management, and ticketing
• Uses Pydantic schemas for input validation and structured request handling
• Sends email notifications securely via environment-configured SMTP
• Includes realistic seeded data for demo and testing workflows

This project demonstrates practical Agentic AI design for enterprise productivity, blending modular tool execution with prompt-based automation.

GenAI Project

AI Knowledge Explorer

Domain/Function: Conversational AI & Information Retrieval

Developed a Streamlit-based AI Knowledge Assistant that retrieves answers from uploaded documents (via ChromaDB embeddings) and real-time web search (SerpAPI). Integrated RAG pipeline, semantic routing, and Groq-hosted LLaMA3 to deliver concise or detailed responses.

View Project
Engagement

View Project
on GitHUB

View Project
on Streamlit

Other Tools

Python

ChromaDB

Groq

SerpAPI

Streamlit

Detailed Mode with Document Analysis

Concise Mode with Web Search

Project Details

Professionals, students, and everyday users often struggle to access reliable, domain-specific information since knowledge is fragmented across documents and the web. To solve this, I built an AI Knowledge Assistant that combines document retrieval with real-time web search.

The system integrates semantic embeddings, ChromaDB-based RAG, and LLaMA3 (via GROQ) to deliver accurate, context-aware answers through a simple Streamlit interface.

Key Features:
• Document ingestion, chunking, and embedding with Sentence Transformers in ChromaDB.
• Real-time web search via SerpAPI when document context is insufficient.
• Hybrid RAG pipeline powered by LLaMA3 for natural language Q&A.
• Flexible response modes: concise summaries or detailed explanations.

This project showcases GenAI in information retrieval and conversational AI, with skills in semantic search, LLM integration, and hybrid knowledge systems.