Projects
Time Series Price Prediction and Feature Engineering with PySpark
Description: Developed a robust time series forecasting system for financial market data using Apache Spark’s distributed computing capabilities.
Key Features: - Implemented advanced feature engineering techniques including lag features, rolling statistics, and technical indicators - Utilized PySpark for scalable data processing of large financial datasets - Applied ensemble methods combining ARIMA, LSTM, and Random Forest models - Achieved 15% improvement in prediction accuracy compared to baseline models
Technologies: PySpark, Python, Pandas, Scikit-learn, TensorFlow
Reinforcement Learning for Derivatives Pricing and Hedging
Description: Built an innovative RL-based system for automated derivatives pricing and dynamic hedging strategies in volatile markets.
Key Features: - Implemented Deep Q-Network (DQN) and Actor-Critic algorithms for optimal hedging decisions - Developed Monte Carlo simulation framework for derivatives valuation - Created dynamic hedging strategies that adapt to market regime changes - Reduced hedging costs by 23% while maintaining risk exposure targets
Technologies: PyTorch, OpenAI Gym, NumPy, Pandas, Quantlab(Financial Mathematics Libraries)
Generative AI Music Model with GANs
Description: Developed a complete generative adversarial network system that converts text descriptions into unique musical compositions.
Key Features: - Built custom GAN architecture with specialized discriminator for musical quality assessment - Implemented text-to-music mapping using transformer-based encoders - Generated 50+ unique music tracks across various genres and moods - Achieved 78% user satisfaction rate in blind listening tests
Technologies: PyTorch, Librosa, transformers, Diffusion Autoencoder
Deep CNN for Text-to-Image Generation
Description: Engineered a convolutional neural network architecture specifically optimized for high-quality text-to-image synthesis.
Key Features: - Designed custom CNN architecture with attention mechanisms and skip connections - Implemented progressive training strategy for stable high-resolution generation - Achieved 30% improvement in image realism scores (FID, IS metrics) - Reduced training time by 40% through architectural optimizations
Technologies: PyTorch, TensorFlow, OpenCV, PIL, CUDA, Distributed Training
Portfolio Optimization with Bloomberg Data and Deep Learning
Description: Developed an advanced portfolio optimization system using real-time Bloomberg market data and wavelet coherence analysis.
Key Features: - Integrated Bloomberg Terminal API for real-time market data feeds - Implemented Wavelet Coherence Analysis (WCA) for multi-scale market correlation detection - Applied deep learning models for dynamic asset allocation optimization - Achieved 18% improvement in Sharpe ratio compared to traditional Markowitz optimization
Technologies: Bloomberg Terminal API, Python, PyWavelets, TensorFlow, Pandas, NumPy
Intelligent Toll Collection System with AI
Description: Built a comprehensive automated toll collection system using computer vision and traffic prediction with graph neural networks.
Key Features: - Developed license plate recognition system using YOLO and OCR technologies - Implemented Graph Neural Networks for traffic flow prediction and congestion forecasting - Created real-time processing pipeline handling 900+ transactions per second - Achieved 76% reduction in manual intervention and processing errors
Technologies: OpenCV, PyTorch, TensorFlow, Graph Neural Networks, Computer Vision, Real-time Systems
AI-Powered Gameplay with Reinforcement Learning
Description: Developed intelligent AI agents for gaming using advanced reinforcement learning techniques and hyperparameter optimization.
Key Features: - Implemented various RL algorithms including PPO, A3C, and DQN - Built automated hyperparameter tuning system using Bayesian optimization - Created adaptive AI that learns and improves through gameplay - Demonstrated superior performance against human players in multiple game scenarios
Technologies: PyTorch, Lua, Python
Demo: YouTube Channel
Full-Stack Banking Website with High-Performance Backend
Description: Engineered a complete banking platform with real-time transaction processing and enterprise-grade scalability.
Key Features: - Built microservices architecture supporting 900+ transactions per second - Implemented real-time transaction processing with ACID compliance - Developed secure authentication and authorization systems - Created responsive frontend with modern UI/UX design principles - Integrated with multiple payment gateways and banking APIs
Technologies: Node.js, React, PostgreSQL, Docker, Kubernetes, Microservices, REST APIs
Note: Each project(repos uploading soon…) represents months of dedicated development, testing, and optimization. Complex real-world problems and continuously learning new approaches to enhance system performance keeps me going.