Projects

Here are some of the projects I've worked on

Computer Vision Based Virtual Mouse Cursor Using Hand Gesture

  • •Developed and deployed a cutting-edge virtual cursor system using Computer Vision technologies; enhanced user engagement metrics indicated a 25% increase in session durations, demonstrating the effectiveness of gesturebased controls.
  • •The hand detected by Hand landmarks 21 segments then gesture mapped to corresponding mouse actions, enabling users to interact with the computer without physical touch
PythonMedipipeOpenCV
Source CodeLive Demo

Spotifyzer : Data warehouse Project

  • •Automated Spotify ETL pipeline, designed a star-schema warehouse (Streaming-Events fact + Tracks, Artists, Audio-Features, Users dimensions) to power sub-second analytical queries. Tech Stack: Python, AWS Glue, Snowflakes, and Power BI.
PythonAWS GlueSnowflakePowerBI
Source Code

RESTful URL Shortener using FastAPI and PostgreSQL

  • •Developed a high-throughput URL shortener with FastAPI, achieving sub-80ms latency for all API endpoints.
  • •Implemented a custom 7-character hashing algorithm, creating a scalable system capable of generating unique URLs.
  • •Utilized PostgreSQL for database management, designing a schema to store original and shortened URLs with their corresponding creation timestamps.
  • •Developed API endpoints for creating, retrieving, and deleting URLs, with comprehensive error handling for invalid inputs and non-existent URLs.
PythonFastAPIPostgreSQLRedis
Source Code

SupportIQ – RAG-Enabled Customer Support Data Pipeline

  • •Developed a scalable data pipeline to ingest, clean, and vectorize customer support tickets enabling LLM-powered Q&A via Retrieval-Augmented Generation (RAG).
  • •Automated the ETL and document embedding process using PySpark and Apache Airflow, deployed via FastAPI and Docker.
  • •Implemented scalable vector search using FAISS and Pinecone, improving response retrieval time and enhancing the accuracy of LLM-generated answers in customer support workflows.
PySparkFastAPIAirflowFAISSPinecone
Source Code

CortexChat – Chat meets Intelligence

  • •Designed a full-stack real-time chat app with secure 1-on-1 messaging, friend requests, and JWT-based authentication, powered by
  • •Socket.IO
  • • and PostgreSQL.
  • •Integrated a RAG-powered AI assistant using LangChain, Pinecone, and LLaMA 3 for contextual document Q&A over chat.
  • •Developed a modular FastAPI backend to handle AI assistant requests, enabling scalable and asynchronous interaction with embedded documents.
  • •Orchestrated a seamless AI pipeline combining vector search, document chunking, and LLM inference, enhancing user experience with relevant, context-aware responses.
ReactNode.jsPySparkLangChainLLaMA3FastAPIPinecone
Source Code