Systems / Networking
DPI Engine - Deep Packet Inspection
Developed as part of a project-based systems programming curriculum, this C++17 deep packet inspection engine analyzes PCAP captures, reconstructs TCP/UDP flows, and classifies application-layer traffic via TLS SNI and HTTP Host inspection. Studied and rebuilt the single-threaded and multi-threaded (load-balancer and fast-path) architectures, with configurable traffic blocking rules.
C++17libpcapTLS/SNIMultithreading
Vector Search Engine
Your Own AI - Vector Database
Developed as part of a project-based systems programming curriculum, this C++ vector database implements HNSW, KD-Tree, and Brute-Force search algorithms across distance metrics with an automated benchmarking suite. Studied and rebuilt a local RAG pipeline via Ollama with document chunking and nomic-embed-text embeddings, exposed through a REST API with a 2D PCA visualization frontend.
C++HNSWKD-TreeRAGOllamacpp-httplib
Fintech / Backend
UPI Mesh - Offline-First Payments
Developed as part of a project-based systems programming curriculum, this Python/FastAPI backend is ported from a Java/Spring Boot reference implementation, reimplementing hybrid RSA-2048/AES-256-GCM cryptography, a simulated Bluetooth mesh network, and ciphertext-hash idempotency for settling UPI-style transactions in zero-connectivity environments. Validated the port with a 27-test pytest suite and CI across Python 3.10-3.12.
PythonFastAPIRSA/AES-GCMSQLiteDockerNginxAWSGitHub Actions
AI / NLP / FastAPI
Hire Me AI - AI Resume Parser & Chatbot
Developed as part of a project-based AI engineering curriculum, this is extended from a resume-parsing exercise into a production FastAPI backend that parses PDF resumes into structured Pydantic schemas and deploys an interactive AI chatbot representing the candidate. Grounded entirely in the actual resume content to prevent hallucination, this system utilizes Groq's high-speed inference engine to deliver fact-bound, professional responses to recruiter queries.
PythonFastAPIPydanticGroqLLM
AI / NLP / RAG
YouTube Scrapper - Multi-Playlist RAG
Developed as part of a project-based AI engineering curriculum, this Retrieval-Augmented Generation (RAG) system is capable of analyzing vast YouTube playlists to extract grounded answers linked to precise video timestamps. Features custom BAAI/bge-m3 embeddings for Hinglish queries, an optimized Qdrant vector database, and fault-tolerant asynchronous batched video transcription utilizing Faster-Whisper.
PythonQdrantFaster-WhisperFastAPIyt-dlp
Computer Vision / AI
Posture Checker - AI Physiotherapy
Developed as part of SIC course curriculum, this AI-driven physiotherapy application provides real-time exercise form feedback using MediaPipe pose estimation. It translates pose landmarks into joint angles for repetition counting and form correction.
PythonStreamlitMediaPipeOpenCV
RAG Application
Conversational RAG Chatbot
Refactored a baseline RAG script into a production-ready web application by securing API key management via environment variables, modularizing the global execution state, and optimizing the dependency tree. Utilizes Streamlit, LangChain, and Groq to facilitate context-aware, multi-turn dialogue over uploaded PDF documents with ChromaDB vector storage.
PythonStreamlitLangChainChromaDBGroqHuggingFace
Computer Vision
Sports Person Classifier
Developed as part of a project-based data science curriculum, this end-to-end computer vision classification system utilizes OpenCV Haar Cascades and Wavelet Transforms for robust feature extraction. Trained and optimized SVM, Random Forest, and Logistic Regression models via GridSearchCV, achieving prediction accuracies exceeding 87%, and deployed via a Flask REST API.
PythonOpenCVScikit-learnFlask
ML Deployment
Home Price Suite
Developed as part of a project-based data science curriculum, this end-to-end machine learning ecosystem delivers a production-grade web application with a Flask REST API, Nginx reverse proxy, and a responsive frontend. The core inference engine utilizes a Scikit-learn Linear Regression model optimized through advanced feature engineering and K-Fold cross-validation techniques.
PythonScikit-learnFlaskNginxDockerAWSGitHub Actions
LLM Pipeline
AI Resume Screener
Developed as part of a project-based AI engineering curriculum, this project extends a script-based resume-scoring exercise into a full Streamlit application, adding structured Pydantic schemas, a batch processing pipeline with caching, and exponential retry-with-backoff around Groq API calls, with an interface that surfaces candidate skill alignments and gaps.
PythonStreamlitGroqPydantic
Web App / ML
House Price Predictor
Developed as part of NIELIT Data Science curriculum, this end-to-end Machine Learning pipeline uses Streamlit and Scikit-learn for property price predictions. It trains a Random Forest Regressor on a real-world dataset of 76,000+ Mumbai property listings, implementing robust data cleaning, categorical encoding, and Train/Test split evaluation metrics (R² and MAE).
PythonStreamlitScikit-learnRandom ForestPandas