Overlay for competitive programming competitions broadcasts

An algorithm for robot navigation was designed, accounting for random obstacles and determining optimal paths. It leverages a genetic algorithm to pinpoint the shortest route from start to end.

Developed TactileNet, the first deep-learning model designed for surface roughness recognition using EEG data. This project leverages CNNs to classify surface textures encountered through a robotic…

The optimization of Wireless Sensor Networks (WSNs) using low-power nodes focused on energy efficiency, introducing a routing strategy for stable nodes based on Ant Colony Optimization (ACO).

Developed a Windows-based app for analyzing data distributions and identifying the best-fitted distribution using the Maximum Likelihood Estimation algorithm. The app features histogram analysis, e…

Developed a Windows tool using PyQt5, integrating K-means clustering for data analysis. The application recommends optimal cluster numbers, identifies cluster members, and allows exporting results …

Autoencoders (AEs) were developed for high-resolution image reconstruction from TinyImageNet and pair recovery from averaged CIFAR-10 composites, demonstrating proficiency in neural network design …

Implementing a Graph Neural Network in PyTorch for edge prediction on MNIST images, showcasing a novel approach in connectivity analysis.

Projects with Genetic Algorithm (GA), Ant Colony Optimization (ACO), and Particle Swarm Optimization (PSO)

Implemented a CNN-LSTM Action Recognizer for dynamic motion analysis, integrating convolutional and recurrent neural networks to efficiently recognize and classify actions in video data of UCF101 d…

Developed the Neuro Vision Transformer, an innovative machine learning model utilizing Vision Transformer architecture for EEG signal classification, focusing on distinguishing ADHD from healthy pa…

Developed a custom application of the Segment Anything Model (SAM) for breast cancer tissue segmentation, utilizing Hugging Face's Transformers and fine-tuning the decoder to predict segmentation m…

Using DIgSILENT, a smart-grid case study was designed for data collection, followed by feature extraction using FFT and DWT. Post-extraction, feature selection. CNN-based and extensive machine lear…

Developed the ViViT model for medical video classification, enhancing 3D organ image analysis using transformer-based architectures.

ADHDeepNet is a model that integrates temporal and spatial characterization, attention modules, and explainability techniques, optimized for EEG data ADAD diagnosis. Neural Architecture Search (NAS…

A 3D Attention U-Net model is developed, aimed at segmenting and tracking Multiple Sclerosis lesions in MRI images.

Developed an AI-driven project for Printed Circuit Board (PCB) analysis, incorporating computer vision for image registration, IC detection, and recognition, along with web scraping for data extrac…

Developed BERT, LSTM, TFIDF, and Word2Vec models to analyze social media data, extracting service aspects and sentiments from a custom dataset. Provided actionable insights to telecom operators for…

Developed a churn prediction model using XGBoost, with comprehensive data preprocessing and hyperparameter tuning. Applied SHAP for feature importance analysis, leading to actionable business insig…

Developed a reinforcement learning framework using Deep Q-Networks (DQN) to optimize scheduling in Wireless Sensor Networks (WSN), enhancing energy efficiency and state estimation through a custom …