I'm passionate about artificial intelligence and machine learning, particularly in areas like computer vision, deep learning, and pattern recognition. I enjoy building intelligent systems that can understand and process visual and textual information, and I’m always exploring how AI can be applied to solve practical, real-world problems.
Responsibilities:
Responsibilities:
A fully automated deep learning pipeline for dog identification via nose patterns. It uses head pose estimation, YOLOv8 for nose detection, and a ResNeSt-based model to generate embeddings. The framework achieves 91.58% AUC and is intended for real-world use in pet recovery and animal services.
Journal Website Google ScholarAn unsupervised deep learning approach using Denoising and Sparse Autoencoders for anomaly detection in NIDS. Evaluated on the NSL-KDD dataset, the system achieved 94.35% accuracy and demonstrated strong capabilities in detecting zero-day attacks.
Civilica LinkDeveloped a machine learning-based recommender system for women's clothing by combining image and text analysis. Designed and trained deep learning models to detect and extract visual features from garments using Convolutional Neural Networks (CNNs), while simultaneously processing product descriptions using Natural Language Processing (NLP) techniques. The model successfully analyzed clothing features and provided intelligent suggestions to users based on item similarity and description relevance.
In this project, I focused on developing the embedding component of a dog nose-print identification pipeline. I trained a deep learning model based on the ResNeSt architecture to extract compact and discriminative embedding vectors from cropped images of dog noses. These embeddings were later used in a re-identification system designed to recognize individual dogs. The model was optimized for generalization and real-world deployment in scenarios such as pet recovery and animal identification services.
Built a real-time facial landmarking system using Python and ML techniques. Utilized live video feeds for head pose estimation and face boundary detection, enabling precise region tracking for biometric and interactive systems.
GitHub ProfileApplied deep learning models to detect breast tumors in Automated Breast Ultrasound (ABUS) images. Focused on increasing diagnostic support accuracy for medical professionals using AI-based image segmentation.
Fine-tuned a Persian language model to support tasks like translation, NER, summarization, title generation, sentiment analysis, and question answering. Enabled multilingual capabilities in domain-specific texts.
Built automation scripts using Selenium to simulate and test interactions on Chess.com, ensuring UI behavior and game logic consistency for various test cases.
Created a responsive web platform for food ordering with dynamic forms, order tracking, and admin dashboard. Designed using PHP and Bootstrap.
Developed an interactive image processing toolbox in MATLAB for filtering, edge detection, thresholding, and object analysis.
I'm focused on areas of artificial intelligence and machine learning, especially computer vision, deep learning, and embedding-based modeling. I’m interested in building intelligent systems that can extract meaningful patterns from images and text, and applying them in real-world tasks like detection, reidentification, and recommendation.
I also enjoy working on end-to-end pipelines from training and evaluating models to deployment and exploring efficient ways to combine text and image understanding in intelligent applications.