yasirali0 Goto Github PK

• My focus was on video recommendation systems where I performed tasks ranging from data analysis, deep learning-based model development and its dockerization up to Flask-based API development.

• I have programmed in various languages such as C++, C, JavaScript, and MATLAB, at a basic level.
• I am proficient in Python and use it for research and development in machine learning and deep learning.

AI Developer Intern | 03 March 2024 - 30 April 2024 | HAMA Lab Co., Ltd., South Korea

• Video Recommendation System
  • Performed data analysis on video and user data within the database to formulate the research objectives.
  • Researched deep learning-based recommendation systems to select appropriate models and strategies tailored to our data.
  • Implemented data and machine learning pipelines and developed training and inference APIs using the Flask package.
  • Incorporated multi-threading strategy within the inference API to efficiently manage user requests and AI model inference simultaneously.
  • Utilized Docker for containerizing the recommendation system, ensuring portability and scalability of the solution.

Graduate Research Assistant | March 2022 - February 2024 | Information Security & Machine Learning Lab, Gachon University, South Korea

• Research on Federated Learning

  • Conducted research in federated learning, focusing on the detection of poisoning attacks within the federated learning paradigm
  • Developed a federated learning framework using Python, PyTorch, and threading
  • Implemented deep learning models such as AlexNet, VGG16, and ResNet18 as the base models for the federated learning environment, and evaluated them on datasets such as MNIST, CIFAR-10, and CIFAR-100
  • Simulated poisoning attacks and analyzed their impact on the accuracy of federated learning
  • Integrated state-of-the-art poisoning attack defense methods into the codebase for benchmarking purposes
  • Proposed a novel defense method that outperformed the state-of-the-art in terms of poisoning attack detection accuracy
  • Authored a research article currently under review in IEEE Transactions for Computational Social Systems

• Research on Tracing Attackers Over Overlay Networks

  • Collaborated with a colleague on this research project aimed at reducing the execution time and memory consumption of deep learning-based correlation attacks against Tor networks
  • Conducted a thorough survey on deanonymization attacks targeting the Tor overlay network, with a specific focus on deep learning-based correlation attacks
  • Performed an in-depth analysis of the prominent deep learning-based correlation attack, "DeepCoFFEA" identifying two critical issues, high memory consumption and execution time
  • Successfully mitigated memory consumption challenge, reducing consumption from 133GB to 70GB through effective memory deallocation and proactive garbage collection strategies
  • Achieved a seven times reduction in execution time by leveraging GPU processing, facilitated by PyCUDA library.
  • Co-authored a research article in IEEE Access journal, outlining the findings and implemented solutions

Intern | February 2021 - April 2021 | National Center of Artificial Intelligence at UET Peshawar, Pakistan

• Landslide Monitoring and Alert System
  • Collected landslide videos to form a dataset for input into deep learning models
  • Segmented and annotated videos into pre-landslide, landslide, and post-landslide phases by utilizing a custom Python script

Languages 👉 Python (Proficient) | C/C++ (Beginner)

ML/DL Frameworks 👉 PyTorch | Keras | TensorFlow | scikit-learn

Python Libraries 👉 NumPy | OpenCV | Matplotlib | Pandas | scikit-image | Tkinter | sqlite3 | PyCUDA | threading

Development Tools 👉 Visual Studio Code | Jupyter Notebook | Git | GitHub | GitLab | Docker | Flask

AI Workflow Experience 👉 Model development | Model optimization | Dockerization | API development

Operating Systems 👉 Ubuntu | Windows

Soft Skills 👉 Communication | Teamwork | Problem-Solving | Critical Thinking

• Code is available here.