cv

This is a description of the page. You can modify it in '_pages/cv.md'. You can also change or remove the top pdf download button.

Basics

Name Muhammad Farhan
Label AI and Automation Engineer
Email farhanzafrani@gmail.com
Phone (+92) 312 9317210
Url https://farhanzafrani.github.io/cv/
Summary As an AI and Automation Engineer with 1 year of experience, I work on cutting-edge research projects for BMW Group, focusing on shape optimization in structural mechanics and applying advanced deep learning techniques to detect anomalies in diverse simulation data formats. My expertise spans a range of deep learning models, including RNNs, CNNs, LSTMs, GNNs, MeshGraphNets, and Transformers, with data sourced from both SQL (Postgres, MySQL) and NoSQL (MongoDB) databases. Additionally, I develop simulation workflows on SimuSpace, ensuring seamless data processing and analysis.

Work

  • 2023.12 - Present
    AI and Automation Engineer
    SOCO Engineers, Multan
    Optimized simulation processes, reducing design iterations by 15% and accelerating design optimization with surrogate modeling. Developed and deployed a high-throughput FEA automation framework using ANSA, Abaqus, Optistruct, and Python.Collaborated on automation solutions, saving engineering hours, and contributed FEA expertise to autonomous driving system safety.
    • Optimization
    • Automation
  • 2022.09 - 2023.05
    Research Assistant
    Human System Lab, SMME, NUST Islamabad
    Designed and fabricated a Soft Robotic Glove for stroke rehabilitation assisting with EMG data analysis and mechanical modeling. Achieved 92% accuracy on real-time data and 98% on offline data by analyzing machine learning models with Fitt’s Law.Developed a real-time EMG data protocol and controlled the Soft Robotic Glove using Raspberry Pi 4
    • Soft Robotics
    • EMG
  • 2022.05 - 2022.09
    School of Mechanical and Manufacturing Engineerins (SMME), NUST Islamabad
    Machine Learning Intern
    Predicted hand gestures using EMG data and machine learning algorithms, achieving 96% accuracy on a generalized test dataset. Designed a data acquisition protocol for Trigo Delsys and collected EMG data from 15 healthy subjects for ML training. Programmed hand-gesture-controlled games, noting overfitting in boosting algorithms like XGBoost and AdaBoost.

Volunteer

  • Mar 2022 - Present

    Islamabad, Pakistan

    Lead Organizer
    Bzm-e-Paigham
    Participate in their Career Campaign *Make the Diamond shine* to motivate the student about their future goals and studies. Worked with them for the 2-week-long campaign in which we visited 4 indigenous schools of Islamabad.
    • Covered above 400 students group-wise, to educate them about *How to study?*
    • Motivate them to apply for the scholarships available for them in the area to fund their studies
    • Work on their personal skills and guide them to become a better citizens of Pakistan

Education

  • 2019.09 - 2023.06

    Islamabad

    BE Mechanical
    National University of Sciences and Technology (NUST)
    Mechanical Engineering
    • Mechanics of Machines
    • Finite Element Analysis
    • Fluid Dynamics
  • 2017.04 - 2019.06

    Islamabad

    Pre-Engineering
    Ali Trust College
    Intermediate
    • Advanced Physics
    • Calculas
    • Analytical Geometry

Awards

  • 2023.06.05
    NGIRI - Winner
    Ignite, Pakistan
    Our Final Year Project was awarded funding of worth 0.1 million PKR by the government of Pakistan under the IGNITE National Technology Fund program..'
  • 2023.03.05
    FICS-Finalist
    FICS-NUST
    Our start-up idea, “REHABOTICS”, won 1st position in the Idea submission stage of Pakistan’s largest start-up competition, wherein more than 450 ideas were presented in the year 2023.

Certificates

Deep Learning Specialization
deeplearning.ai & Stanford University 2018-01-01

Publications

  • under.Review
    An sEMG-Driven Multiple DOFs Interactive Games Based Rehabilitation Protocol for Stroke Patients: Assessing Real-Time User Performance Feedback
    Heliyon
    This study explores the effectiveness of interactive home-based rehabilitation (HBR) using a 2D video game designed for stroke patients with hemiparetic hands. While video game-based therapy (VGT) is a common method for upper limb rehabilitation, many existing approaches lack real-time feedback and clinical usability. The proposed rehabilitation protocol was tested over three weeks with ten healthy individuals and two chronic stroke patients, measuring performance through Fitt’s law. Significant improvements were observed in stroke patients, with increased scores, completion rates, and throughput over time, suggesting that the protocol holds promise for effective stroke rehabilitation.
  • NOT IN SCOPE
    Development of Raspberry Pi Based sEMG Driven Standalone Upper Limb Rehabilitation Training System: A Human-Computer Interaction
    IEEE Conference on Future Technologies
    This study introduces a new approach to stroke rehabilitation by turning traditional physiotherapy exercises into interactive video games. Recognizing that conventional stroke therapy can be repetitive and boring, the researchers developed a system that uses pattern recognition to make the process more engaging for upper limb stroke patients. The system features two games that mimic physiotherapy exercises, designed to improve motor function in the arms. Using a Myo armband, the team collected muscle activity data (sEMG) from ten healthy participants to train a decision tree algorithm. When tested, the system showed promising results, with the algorithm achieving over 91% accuracy in offline tests and nearly 98% accuracy in real-time gameplay. These findings suggest that gamified rehabilitation could be an effective and enjoyable way to help stroke patients regain motor function.

Skills

Machine Learning
Regression Algorithms
Classification Algorithms
Ensemble Learning
Boosting Algorithms
Clustring Algorithms
Deep Learning
Neural Networks
Recurrent Neural Networks
Transformers
Reinforcement Learning
Autoencoders
Generative Adversarial Networks
Graph Neural Networks
Cloud Platforms
AWS
Azure
Kubernetes
Docker
Serverless Computing

Languages

Pashto
Native Language
Urdu
National Language
English
Fluent

Interests

Computational Scieces
Machine Learning
Optimization Techniques
Deep learning
LLMs
Generative AI
Finite Element Analysis

References

Dr. Muhammad Asim Waris
Dr. Muhammad Asim Waris is an Assistant Professor at the Aerospace department SMME, NUST. He was my supervisor for the bachelor's (BE) final year project, and currently, I was working with him as a research assistant on a research project at the Human System Lab, SMME, NUST.
Dr. Ibraheem Haneef
Dr. Ibraheem Haneef is a Professor at the Aerospace department SMME, NUST. He has taught me Project Management and also coordinated me in my final year project. I took a workshop of Practical Reinforcement Learning under his supervision.

Projects

  • 2022.06 - 2023.06
    Design and Fabrication of Soft Robotics Glove for the Rehabilitation of Stroke Patients using EMG data and Pattern Recognition Techniques
    Fabricated a right hand Soft Robotic Glove to assist the Stroke Patients in their Rehabilitation. Casted the Soft Robotic Actuators using Ecoflex 00-30 (silicone) that is an synthesized elastomer used in pneumatic based Soft Robotics. Achieved the desired bending using lower pressure of 60 Kpa. Controlled the actuator using the Raspberry PI 4 that was programmed using Python based on Machine Learning algorithms. Designed the Actuator using CAD software and simulated the actuator using ANSYS workbench for pressure requirements that came out to be 60 Kpa for single Actuator. Achieved an accuracy of 92% in prediction of the true hand gesture that ultimately controlled the actuator to rehabilitate the Stroke Patient Hand through Robotic Glove
    • Soft Robotics
    • EMG
    • Machine Learning