Projects

AI Fitness Trainer– Acharya (Mar ’25)

  • Built an AI-powered fitness platform that generated over 500 personalized workout and diet plans, leveraging Gemini AI for real-time recommendations and achieving high goal alignment.
  • Developed a voice-enabled coaching system using Vapi and 11Labs, supporting 200+ hands-free personalized sessions.
  • Designed a scalable Next.js + React architecture with Convex for state and realtime, supporting 1,000+ concurrent users and 99.9% uptime.
  • Delivered real-time program generation and progress tracking, increasing daily active user engagement and reducing churn.

Real-Time Weapon Detection System with Face Recognition (Sep ’25)

  • Developed a real-time surveillance system that detected weapons and unknown individuals with high accuracy to enhance security monitoring.
  • Implemented a computer vision pipeline with YOLO-based detection and Keras-FaceNet for facial recognition, processing 30+ FPS for live feeds.
  • Automated threat responses (alarms and email notifications within 2 seconds) and built a Flask dashboard supporting 100+ concurrent streams.

Student Feedback Review System (Sep–Dec 2024)

  • Flask web app for analyzing student feedback using NLP
  • Sentiment Analysis & Text Classification (Naïve Bayes, Random Forest, SVM)
  • TF-IDF vectorization, hyperparameter tuning, data visualizations

Redis Clone (C++)

  • Technologies: C++, TCP Sockets, STL, Multithreading, Networking.
  • Built a Redis-like in-memory data store from scratch in C++, supporting 12+ Redis commands and achieving O(1) average-time complexity for key–value operations using unordered_map-based storage; validated across thousands of read/write operations.
  • Implemented core Redis commands including PING, ECHO, SET/GET with PX expiry, and List operations (RPUSH, LPUSH, LRANGE, LLEN, LPOP), handling both positive and negative indexing and supporting range queries over lists of 10⁴+ elements.
  • Designed and implemented blocking commands (BLPOP) with FIFO-based fairness, correctly serving multiple simultaneously blocked clients (5–10+), supporting indefinite blocking and timeout-based unblocking with second-level precision.
  • Managed concurrent TCP client connections with multithreading, ensuring thread-safe command execution, efficient parsing of hundreds of sequential commands per session, and structured response serialization inspired by Redis protocol semantics.
  • Optimized for low latency and correctness: millisecond-level key expiration with zero stale reads after expiry; deepened expertise in network programming, data structures, synchronization primitives, and event-driven server design.

Open Source Virtual Memory Manager in MINIX 3 (Feb–May 2025)

  • Implemented FIFO, LRU, LFU, Second-Chance page replacement
  • Kernel-level debugging, Unix OS internals

Web-based Mini C Compiler

  • Technologies: Lex, Bison, C, Python (Flask), HTML, CSS, JavaScript
  • Built a web-based Mini C Compiler using Lex & Bison for parsing and IR generation, with a Flask/C backend and a modern web frontend.
  • Enabled code editing, real-time output, and modular compiler phases—bridging low-level tools with web tech.