Portfolio - Vraj Patel

GenAI Project

HR Assist

Domain/Function: Agentic AI for Human Resources

Built an AI-driven agentic system to automate HR onboarding workflows. Integrated with Claude Desktop via MCP for seamless task orchestration, added secure credential and email automation, and designed a scalable architecture extendable to payroll and compliance.

View Project
on GitHUB

AI agent automating HR onboarding by adding employees, verifying managers, and updating records seamlessly.

AI agent executing end-to-end HR onboarding tasks—sending emails, notifying managers, and raising equipment tickets.

Project Details

Engineered a credit scoring framework that lifted AUC from 82% to 88% and attained a KS statistic of 46, exceeding
industry thresholds.
Automated feature generation, scaling, and outlier treatment, reducing preparation workload by 35%.
Refined predictive capability through Optuna and RandomizedSearchCV, elevating the Gini coefficient from 78 to 85.
Supplied an interactive Streamlit interface for analysts, enabling dynamic applicant reviews and lowering projected
defaults by 12%.

GenAI Project

E-Commerce Chatbot

Domain/Function: GenAI RAG for Retail Platforms

Built a GenAI-powered e-commerce chatbot with RAG for real-time product queries and FAQs. Integrated LLaMA 3.3 + GROQ for fast responses and delivered a Streamlit interface with history and product filters to improve customer support efficiency.

View Project
on GitHUB

GenAI-powered e-commerce chatbot fetching real-time product data and handling customer queries with accuracy and speed.

E-commerce chatbot retrieving real-time Puma shoe deals with discounts, ratings, and product links from live databases.

Project Details

Built a GenAI chatbot for e-commerce platforms to handle customer queries and product discovery.
Implemented RAG pipeline with SQL database integration for real-time pricing, availability, and policy updates.
Enabled intent recognition to distinguish between FAQs and database-driven queries.
Optimized performance using LLaMA 3.3 with GROQ API, cutting response latency and improving user experience.
Designed a Streamlit interface with conversational history and product filtering for intuitive customer interaction.
Delivered a 3× faster query resolution compared to static FAQ systems, improving customer satisfaction.
Built the system to be scalable and adaptable across retail platforms.

Machine Learning Project

Credit Risk Modeling

Domain/Function: Finance and Banking

Developed a credit scoring system to assess borrower risk using machine learning. Automated data preparation, optimized predictive models, and built an interactive Streamlit tool for analysts to review applicants and improve lending decisions.

View Project
on GitHUB

App takes inputs like income, loan amount, tenure, and ratios to calculate default probability, credit score, and risk rating.

Credit risk modelling app predicting default probability, credit score, and rating based on borrower financial inputs.

Project Details

Engineered a credit scoring framework that lifted AUC from 82% to 88% and attained a KS statistic of 46, exceeding industry
thresholds.
Automated feature generation, scaling, and outlier treatment, reducing preparation workload by 35%.
Refined predictive capability through Optuna and RandomizedSearchCV, elevating the Gini coefficient from 78 to 85.
Supplied an interactive Streamlit interface for analysts, enabling dynamic applicant reviews and lowering projected defaults by 12%.

Deep Learning Project

Car Damage Detection

Domain/Function: Computer Vision, Automotive AI

Built a computer vision pipeline with EfficientNet + YOLO for defect detection. Automated data prep, added heatmap visualizations in Streamlit, and improved insurance claim assessments by speeding up inspection cycles and boosting efficiency.

View Project
on GitHUB

AI-powered Vehicle Damage Detection system identifying front-end breakage from uploaded images for automated assessment.

AI-powered Vehicle Damage Detection system identifying rear-end breakage from uploaded images for automated assessment.

Project Details

Crafted a computer vision pipeline with EfficientNet and YOLO, attaining 92% accuracy in defect localization.
Constructed a data augmentation workflow (resizing, balance correction, random transformations) that trimmed dataset prep time by
40%.
Produced severity-based heatmap overlays in Streamlit for intuitive visualization of damaged surfaces.
Accelerated insurance assessments by shortening inspection cycles by 30%, improving claim efficiency.

Python Project

Expense-Management-System

Domain/Function: Personal Finance & Budget Analytics

Developed a modular full-stack app with FastAPI backend and Streamlit frontend for real-time expense tracking. Added automated testing, RESTful APIs, and async handling to ensure scalability, reliability, and seamless data synchronization.

View Project
on GitHUB

Expense Tracker app interface to add, update, and categorize daily spending with notes for better financial management

Expense Tracker analytics dashboard showing category-wise spending breakdown with charts and percentage insights.

Project Details

Architected a modular full-stack application with separate frontend and backend components for scalable deployment.
Implemented FastAPI-based RESTful backend server with asynchronous request handling for optimal performance.
Designed interactive Streamlit frontend interface providing real-time expense tracking and visualization capabilities.
Developed automated testing suite covering both frontend and backend components to ensure application reliability.
Utilized Python's modern web frameworks to create responsive user experience with real-time data synchronization.
Implemented RESTful API architecture enabling seamless communication between frontend and backend services

Hello, I am

Vraj Patel

2

1

1

1

About

Key

My

HR Assist

E-Commerce Chatbot

Credit Risk Modeling

Car Damage Detection

Expense-Management-System

Awards

Contact Me

Email

Github

Linkedin

Call

Vraj Patel

Download Resume