A full-stack comparison web application built with React.js and FastAPI, featuring AI integration through LangChain and OpenAI API for intelligent, structured comparisons. It combines robust frontend and backend engineering, API design, testing, and prompt-driven AI reasoning into a cohesive, reliable system.

Software Engineering SDLC project covering requirements (FRs & NFRs), design and analysis (UML diagrams: Use Case, Class, Domain Model, and Sequence), full-stack implementation (Firebase backend database and hosting + Flutter frontend), testing (plans and cases), and deployment of a smart laundromat web system.

An automated Smart Toll Gate System built using Raspberry Pi, integrating sensors and RFID for vehicle detection and identification, along with speed and weight measurement for dynamic toll calculation. The system controls gate access, uploads data to ThingSpeak for cloud monitoring, and supports operator overrides and alert mechanisms. It also includes RESTful Flask APIs that enable web-based gate operations, real-time vehicle information, and trespassing event handling through defined routes.

Java-based multithreaded RPG engine showcasing professional software architecture with advanced objected oriented programming. Implements six software design patterns Observer, Singleton, Template Method, Strategy, State, and Command. Uses concurrent threads for battle timing and real-time gyroscope, accelerometer, and sound input with synchronized, thread-safe execution.

Full-stack AIoT engineering project integrating embedded hardware, cloud infrastructure, and machine learning for real-time air quality monitoring and forecasting. Developed solar-powered ESP32 sensor nodes, a Firebase–GCP cloud pipeline, and a Flutter dashboard for live data visualization and predictive analytics using a hybrid CNN–BiLSTM model.

A full-stack AI comparison web application built with FastAPI, React.js, LangChain, and GPT-4o, enhanced with Brave Search grounding for accurate, real-time context. Supports personalized, category-aware comparisons, schema-validated outputs, caching, and PostgreSQL-backed history with follow-up reasoning and shareable comparison links.

On-device fall detection system for elderly care, running entirely on an Arduino Nano 33 BLE Sense with no phone or cloud dependency. Designed a selective multi-stage inference cascade that resolves everyday activity with lightweight arithmetic and only wakes the neural network when a fall is likely, achieving 97.5% live accuracy while cutting per-event energy by 98.4% versus an always-on CNN. Built the full pipeline from IMU data collection and windowing to threshold derivation, logistic-regression gating, CNN training, and TensorFlow Lite deployment as a 5.86 KB model on a Cortex-M4 board.

Semi-supervised machine learning framework for datasets with limited labels, demonstrating applied AI/ML and data science skills in predictive modeling. Includes rigorous data preprocessing, selective pseudo-labeling, and a three-phase training workflow, achieving strong result of test F1 = 0.9986, surpassing published benchmarks

Developed a supervised learning framework to predict protein stability upon single-point mutations using thermodynamic data derived from trypsin and chymotrypsin experiments. Implemented KNN, SVM, Decision Tree, and Naïve Bayes models with a rigorous pipeline including outlier removal (Z-score), standardization, class rebalancing (hybrid over- and undersampling), feature selection (SelectKBest), and dimensionality reduction (PCA). Performed 10-fold cross-validation and hyperparameter tuning across models, achieving strong generalization and robust predictive accuracy. Evaluated performance using F1-score, precision, recall, confusion matrix, and AUC-ROC curves to ensure stability predictions were reliable across all mutation classes

Python Java C++ JavaScript SQL HTML

FastAPI Flask REST APIs OpenAI API Firebase API Google Maps API

React.js Flutter HTML TailwindCSS

• Frameworks: scikit-learn LangChain OpenAI API TensorFlow/Keras TensorFlow Lite
• Techniques: Supervised & Unsupervised Learning, Feature Engineering, Dimensionality Reduction (PCA), Model Evaluation (AUC, ROC), Hyperparameter Tuning, Prompt Engineering, TinyML & Edge AI Deployment, Model Quantization, CNNs

Matplotlib Seaborn Power BI

SQL (Oracle) NoSQL (Firebase Realtime Database)

Google Cloud Platform (GCP) Firebase Hosting Microsoft Azure