Projects

• AI-Driven Legal Research Engine for Courts
  • Backend Development: Built a robust backend using FastAPI to enable efficient API handling.
  • AI Integration: Incorporated Llama 3.1 with Retrieval-Augmented Generation (RAG) to provide contextually relevant insights.
  • Dataset Curation: Curated a custom dataset of 1000+ case summaries to enhance model training and accuracy.
• Cross-Platform AI Assistant
  • Deployment: Developed a cross-platform assistant using Flutter for desktop, Android, and web platforms.
  • AI and Tools: Integrated Gemma2, Llama3, and Mixtral8 LLM models with Groq API for faster inference and ChromaDB for embedding generation.
  • Backend Development: Designed a FastAPI-based backend utilizing RAG for efficient data retrieval.
• Invoice Pattern Matching System
  • AI Model: Implemented TF-IDF vectorizer to convert invoice text into vectors and used cosine similarity with argmax to identify the most similar match.
  • Self-Learning Algorithm: Developed a feedback-driven retraining system to improve model accuracy dynamically.
  • Online Learning: Integrated PassiveAggressiveClassifier for continuous learning from user feedback.
  • Accuracy: Achieved a 96% accuracy score in pattern matching.
• System Monitoring Tool
  • Cross-Platform Development: Built using Flutter for Windows, Linux, and macOS to track system resource usage.
  • AI Integration: Integrated DeepSeek-r1 using Groq API to predict bottlenecks, suggest optimizations, and generate system performance summaries.
  • Real-Time Monitoring: Utilized PowerShell scripts for efficient tracking of CPU, RAM, disk, and network usage.
  • Performance Optimization: Optimized system calls to ensure low-latency monitoring and minimal resource consumption.
• Class Clockwise
  • Timetable Management: Developed a Flutter app to display daily university lecture schedules dynamically.
  • Google Sheets API: Integrated AppScript API to fetch real-time data directly from Google Sheets.
  • Multi-Platform Deployment: Deployed the application on web and Android for seamless access.
• VIIT Chatbot
  • Fine-Tuning & Optimization: Fine-tuned TinyLlama with 4-bit quantization for efficient processing.
  • Retrieval-Augmented Generation (RAG): Implemented Baai embeddings for enhanced response accuracy and real-time updates.
  • Deployment: Deployed the chatbot on Vercel, using MongoDB for storing up-to-date data.
  • Research Contribution: Published findings in IJRESM journal.
• Taxes with LLaMA (Contributed)
  • AI-Powered Tax Optimization: Contributed to an AI-based tax-saving assistant that analyzes income, investments, and financial data to provide tax optimization recommendations.
  • Personalized Insights: Utilized LLaMA models to generate tailored tax-saving strategies for users.
• Aerovania App
  • Full-Stack Development: Led the development of a Flutter-based e-commerce and e-learning app.
  • Backend Integration: Used Firebase for real-time database management, authentication, and cloud storage.
  • User Experience: Designed an interactive UI to enhance usability and engagement.
• ElderCare
  • Healthcare & AI Integration: Built a healthcare platform for elderly care, connecting patients, doctors, and physiotherapists via an Android application.
  • Role-Based Access: Developed an admin panel to assign registered doctors to elderly patients for personalized care.
  • Real-World Impact: Successfully received positive feedback from healthcare professionals and elderly users.
• AirChat
  • Real-Time Communication: Built a basic chat application using Flutter with support for real-time messaging and photo sharing.
  • Backend Services: Integrated Firebase for user authentication, message storage, and notifications.
• Portfolio App
  • Personal Branding: Developed a Flutter-based portfolio application showcasing personal achievements, projects, and skills.
  • Interactive UI: Designed a smooth, user-friendly experience for recruiters and visitors.
• Deception Detection
  • AI for Cybersecurity: Developed a phishing website detection system using DistilRoBERTa, a pre-trained text classification model from Hugging Face.
  • Backend Development: Built the detection API using Flask for fast response times.
  • Chrome Extension: Created a browser extension using HTML, CSS, and JavaScript for real-time phishing detection.
• Replichat Chrome Extension
  • Automated Messaging: Developed a Chrome extension using GPT-3 to generate polite and context-aware replies.
  • Multi-Platform Support: Enabled WhatsApp and email automation for quick and efficient responses.
  • Frontend Development: Built using HTML, CSS, and JavaScript for seamless browser integration.
• Car Resale Price Prediction
  • Web Application: Developed a car resale price prediction website using HTML, CSS, and JavaScript.
  • Machine Learning Model: Used LGBMRegressor and RandomForestRegressor to predict car resale prices with 91% accuracy.
  • Data Analysis: Performed Exploratory Data Analysis (EDA) to identify trends and patterns in the dataset.
  • Backend API: Built the backend using Flask for model inference and API deployment.