Developed an innovative framework for financial time series prediction, integrating advanced deep learning techniques with scalable architecture.

• Designed a hybrid model combining CNNs with multi-head attention at each layer, Bidirectional LSTMs, and fully connected layers for robust feature extraction and temporal modeling.
• Implemented innovations like Mix of Experts, residual blocks, and advanced optimizers (e.g., AdamW) for enhanced generalization and faster convergence.
• Engineered modular pipelines for data preprocessing, model training, evaluation, and result visualization, ensuring reproducibility.
• Enhanced evaluation with metrics (MAE, MSE, RMSE, R²) and detailed plots (error distributions, prediction trends).
• Achieved significant accuracy improvements, reducing RMSE by 15% over baseline models.

Technologies & Tools: TensorFlow, Keras, Python, Matplotlib, NumPy, pandas, scikit-learn, JSON for model serialization
Methodologies: Convolutional Neural Networks (CNNs) with Multi-Head Attention per Layer, Bidirectional Long Short-Term Memory (BiLSTM), Fully Connected Layers, Residual Networks, Mix of Experts, Early Stopping, AdamW Optimizer, Model Evaluation Metrics, Sequence Modeling, Hyperparameter Optimization

• Conceived and initiated the integration of Whale Optimization Algorithm (WOA) with HEFT (Heterogeneous Earliest Finish Time) scheduling to create a pioneering hybrid algorithm aimed at optimizing cloud computing environments.
• Developed and implemented advanced bio-inspired strategies, utilizing the foraging behaviors of humpback whales to significantly improve task allocation and resource scheduling.
• Engineered a simulation environment to rigorously test and evaluate the effectiveness of the WOA-HEFT hybrid algorithm under various operational conditions.
• Analyzed data to extract meaningful insights, leading to substantial improvements in processing speeds and resource efficiency.
• Results will be presented at major academic forums, demonstrating the algorithm’s potential to enhance cloud computing operations and its applicability in real-world scenarios.

• Investigated foundational principles of three main types of recommender systems.
• Developed advanced preprocessing techniques that increased data processing speeds by 300 times.
• Integrated Apache Spark, multiplying processing
• gains tenfold.
• Achieved a cumulative 3000-fold increase in system efficiency.

• for the agricultural industry using Django, REST APIs, and SQL Server, enhancing data management and security.   

were optimized to develop a high-precision image panorama tool. This project, concluding a series of ten, successfully demonstrated the capability to seamlessly integrate multiple images into panoramic views through rigorous testing and team collaboration.

1. I developed a personalized movie recommendation system using the K-Nearest Neighbors (KNN) algorithm. This project focused on accurately aligning movie suggestions with individual use preferences by effectively analyzing user data. The system was crafted to provide precise and relevant movie recommendations, illustrating the practical application of KNN in academic settings.