Rahul Babu

Live Demo Code

• Built and deployed an interactive 3D portfolio experience using Next.js and TypeScript with Three.js and React Three Fiber (Drei, Rapier) for immersive UI and animations.
• Integrated Hygraph (GraphCMS) using GraphQL and Apollo Client to manage portfolio content through a CMS-driven workflow.

Live Demo Code

• Deployed a Retrieval-Augmented Generation application that retrieves relevant policy documents and generates context-enriched responses for climate action recommendations.
• Built an interactive Streamlit chat interface with multi-session history so users can manage and review conversations.

Live Demo Code

• Developed an interactive web application to visualize Arizona opioid-risk data at the Census Block Group level using a FastAPI backend and React + TypeScript frontend.
• Designed backend filtering and API serving and packaged the system with a unified Dockerfile for streamlined deployment.

Developed a fake job posting predictor to identify fraudulent listings that mimic legitimate roles and help improve trust and safety for job seekers and hiring platforms.

Code

• Built a machine learning workflow to classify job posts as real or fraudulent using textual and structured listing signals.
• Framed the system for real-world safety use cases, such as flagging suspicious listings before applicants share personal information or make payments.

Python

Machine Learning

NLP

Fraud Detection

Classification

Code

• Designed a distributed NoSQL database system using Kafka, Redis, and MongoDB, achieving 100% consistency post-recovery and under 500 ms average response time for e-commerce data operations in fault-tolerant scenarios.
• Optimized cross-region replication with FastAPI middleware, ensuring zero data loss during node failures and 248 ms response times during outages through Kafka-backed asynchronous recovery.

Kafka

Redis

MongoDB

FastAPI

Distributed Systems

NoSQL

Fault Tolerance

Cross-Region Replication

Led a research project as the team lead for an 8-member group and documented a 60-page research paper on user identification, authentication, and interaction analysis for social media data analytics using AI.

Code

• Directed project planning, research coordination, and technical execution across an 8-member team.
• Compiled and authored a comprehensive 60-page technical research report covering system design, experimentation, and findings.

Artificial Intelligence

Social Media Analytics

User Authentication

Research Leadership

Technical Documentation

Built a configurable movie-review search experience using Elastic Search UI with App Search engine integration, designed for quick setup, customization, and static deployment.

Code

• Implemented engine-driven search configuration via JSON (`engine.json`) for fields, facets, sorting, and query suggestions without regenerating the app.
• Set up reproducible local development with Node 16.13.0 and npm, and enabled static asset build-and-embed workflow for production hosting.

React

Elastic Search UI

Elastic App Search

Node.js

npm

JSON Configuration

Static Build

Netlify Deployment

Built a web automation and evaluation pipeline to run prompts through multiple LLMs, aggregate responses in structured sheets, and benchmark output quality with NLP metrics.

Code

• Automated prompt execution with Selenium IDE across ChatGPT, Gemini, Claude, Perplexity, Meta AI, and HuggingChat using a localhost prompt-reader workflow.
• Captured model responses in Google Forms and exported to Excel/CSV for downstream analysis and reporting.
• Implemented Python-based evaluation workflows (Response_Evaluation.ipynb) using BLEU, ROUGE, cosine similarity, Distinct-1/2, sentiment, response-length analysis, and word clouds.
• Included data-scraping support for 790 climate policy articles from c40knowledgehub.org to build evaluation and comparison datasets.

Selenium IDE

Python

LLM Benchmarking

BLEU

ROUGE

Cosine Similarity

Distinct-1/2

Data Automation

Code

• Implemented SVM, Logistic Regression, Naive Bayes, XGBoost, BiLSTM, and CNN models on 416K tweets, achieving 97% accuracy with CNN using TF-IDF/word-embedding preprocessing and stratified sampling.
• Identified SVM as the most robust model (94.15% accuracy) through comparative ROC and confusion matrix analysis.

Python

Amazon SageMaker

NLP

SVM

Logistic Regression

Naive Bayes

XGBoost

BiLSTM

CNN

TF-IDF

This project focuses on establishing a robust and secure framework for the management of IoT devices, underpinned by the principles of data immutability and transparency, all facilitated by the innovative use of blockchain technology.

Code

• Implemented a blockchain-based solution for IoT device management.
• Recorded critical transactions including device registration, management, and ownership transfers.
• Developed a user-friendly web application using ReactJS.

MetaMask

Truffle Framework

Ganache

Solidity

React.js

This project focuses on utilizing hyperspectral imaging for forest terrain monitoring, aiming to provide detailed insights into forest contents and potential risks. It utilizes the Indian Pines dataset, which captures hyperspectral data from a forested area, and employs a hybrid classification method, including Support Vector Machines (SVM), 2D Convolutional Neural Networks (CNN), 3D CNN, M3D-CNN, SSRN, and a Hybrid model for precise forest analysis.

Code

• Utilized Indian Pines dataset for hyperspectral imaging.
• Employed hybrid classification methods combining SVM, 2D CNN, 3D CNN, M3D-CNN, SSRN, and a Hybrid model.
• Provided comprehensive insights including vegetation cover percentage, building presence, and hazard detection.

Scikit-Learn

Machine Learning

TensorFlow

Exploratory Data Analysis

NumPy

Data Science

Data Analysis

Python

Pandas

The Web Phishing Detection project is a Python-based solution deployed on the IBM Cloud platform, focused on the detection of phishing websites. It incorporates machine learning techniques to enhance accuracy.

Code

• Trained 10 machine learning models for phishing detection.
• Deployed the highest accuracy model on IBM Cloud.
• Developed a website interface for user interaction with the phishing detection system.

Machine Learning

Exploratory Data Analysis

Data Science

Data Analysis

Python

An application developed using Kivy and PyTorch Framework. The application simulates an autonomous vehicle navigating a map I designed, complete with manually placed obstacles.

Code

• Used Deep Q Networks (DQNs) with relu activation functions for navigation.
• Designed a reward system to improve obstacle avoidance capabilities.
• Received approval for publication in ADCOM Volume 133.

Machine Learning

NumPy

Python

The Tournament Manager is a Java-based application designed to facilitate the efficient organization of sports tournaments.

Code

• Designed and configured tournaments for different divisions and sports.
• Streamlined team enrollment and player information updates.
• Developed using Java and MySQL.

Java

The "Harmful Comments Classifier" is a Python-based Natural Language Processing (NLP) project, executed within the Google Colab environment.

Code

• Developed a deep learning model for classifying comments based on toxicity.
• Utilized embedding techniques, bi-directional LSTM, and dense models.
• Executed within the Google Colab environment.

Machine Learning

Exploratory Data Analysis

Data Science

Data Analysis

Python

"EATERIO" is a web application built using HTML, CSS, JavaScript, and MongoDB, designed to facilitate food ordering and management within organizations.

Code

• Streamlined the food ordering process within organizations.
• Developed using HTML, CSS, JavaScript, and MongoDB.
• Provided a user-friendly interface for placing and managing food orders.

HTML

CSS

JavaScript

MongoDB