My Projects

CIGIN - Chemical Interpretable Graph Interaction Networks

  • Enhanced graph neural network architecture for chemical property prediction with improved stability and interpretability
  • Implemented advanced fault tolerance mechanisms including gradient clipping, loss normalization, and comprehensive error handling
  • Developed transformer-based aggregation module with configurable attention heads to outperform traditional Set2Set pooling
  • Restructured codebase following PEP-8 standards with rich-style logging and JSON structured outputs for better debugging
PyTorch DGL Graph Neural Networks Transformer Architecture Chemical Informatics Python

Molecular Dynamics Analysis of Protein 1a62_A

  • Conducted all-atom molecular dynamics simulations on protein 1a62_A utilizing Dynamic PDB dataset
  • Extracted and analyzed atomic coordinates, velocities, and forces using Python libraries (MDAnalysis, NumPy)
  • Employed Linux-based high-performance computing environments for data processing tasks
MDAnalysis NumPy Python HPC Linux Computational Chemistry

AI-Driven Glycemic Risk Prediction and Health Management System

  • Built ML model with 88% accuracy for diabetes risk prediction from patient records
  • Integrated Gemini AI for personalized lifestyle recommendations, increasing user engagement by 50%
  • Developed a Gradient Boosting Classifier for diabetes risk prediction with high accuracy
Python Machine Learning Gemini AI Healthcare AI Gradient Boosting

Video demonstration of the AI-Driven Glycemic Risk Prediction project

AI-Powered Forex Trading Dashboard

  • Developed end-to-end AI trading system analyzing forex market patterns with 62% accuracy
  • Implemented 20+ technical indicators with Pandas/NumPy for comprehensive signal validation
  • Created an intuitive dashboard interface for real-time market analysis and trading signals
Python Pandas NumPy Machine Learning Financial Analysis

Japanese to English Neural Machine Translation

  • Developed Flask-based translation application achieving 32 BLEU score for Japanese-English translations
  • Optimized inference speed by 35% through quantization and multi-threading techniques
  • Built a modular project structure with potential for cloud deployment
Python Flask MBART NLP HuggingFace

License Plate Detection and Recognition

  • Implemented a computer vision-based model to detect license plates from car images
  • Used OCR techniques to extract and recognize text from the detected plates
  • Built a user-friendly interface for real-time plate detection and recognition
Python Computer Vision OCR Deep Learning

Japanese-to-English Video Dubbing Using BERT and Open Voice

  • Developed an automated system for translating and dubbing Japanese videos into English
  • Used BERT for high-quality translations and Open Voice for natural-sounding English dubbing
  • Research paper accepted at ICMED 2025 conference
BERT Open Voice NLP Speech Synthesis Machine Translation

Pest Detection in Crops

  • Analyzed 19,404 rice pest entries, identifying patterns through visualizations and reducing the dataset to 9,126 entries
  • Preprocessed data for machine learning models (Logistic Regression, SVM, AdaBoost, Random Forest) to achieve high accuracy in pest detection
  • Developed a system to help farmers identify and address pest infestations early
Python Machine Learning Data Analysis Agricultural Technology