Fateme Bafghi

I'm

I am Fateme

Results‑driven Data Scientist with over 6 years of hands‑on experience. Proficient in designing, developing, and integrating AI applications using PyTorch, Keras, and TensorFlow for development, and Docker, TensorRT, and ONNX for deployment. Successfully delivered more than 10 impactful AI projects in collaborative team settings, demonstrating precision and creativity. Recognized for achieving project milestones and adapting strategies to meet evolving requirements. Thrives on embracing and overcoming complex challenges. Committed to continual professional development, evidenced by staying current with the latest advancements in the field.

AI Researcher and Python Developer

  • Birthday: 21 Jan 1994
  • Phone: +989215294507
  • City: Tehran, Iran
  • Degree: Master in Artificial Intelligence
  • Email : fateme.bafghi1994@gmail.com
  • Freelance: Available

for more information please have a look at my cv

My experiences

Education

M.Sc. Artificial Intelligence

September 2016 - September 2019

University of Isfahan, Isfahan, Iran

B.Sc. Computer Engineering, Software Engineering

September 2012 - September 2016

Isfahan University of Technology, Isfahan, Iran

Professional Experience

Data Scientist

May 2023 - Present

Azki, Tehran,Iran

Computer Vision Engineer

May 2023 - Present

Nojan Co, Tehran,Iran

  • Improved the utilized models by extracting most informative data from dataset using different Data Pruning methods
  • Pistachio Classification Model:Designed and developed an innovative Pistachio Classification system utilizing Vision Transformer networks
  • Dataset Subset Clustering: Implemented a clustering method using self-supervised training in ConvnextV2 and Clip model in order to cluster each class of dataset into multiple subclasses

Machine Learning Engineer

JUL 2022 - May 2023

Revivoto, Tehran,Iran

  • Optimizing Models: Achieved impressive results by deploying and optimizing Image-Enhancement, Scene Classification, and Object Removal networks. Reduced model size and memory consumption by 70%, and also Decreased processing time by 50% while improving accuracy.
  • Serving and Calibrating Models: Serving multiple models efficently using TensorRT by calibrating and quantizing models.
  • Data Engine: Developed and designed an efficient Data Engine that streamlined the dataflow, development, and deployment process, utilizing MongoDB to manage and optimize large datasets
  • Scene Classification: Designed and developed advanced Scene Classification networks utilizing transformer networks and data pruning

AI Developer

March 2020 till now

AmeraAndish, Tehran,Iran

  • Developing first automatic answering machine based on AI to answer phones in Persian
  • Developing first analyzer system for Iranian call centers based on their data packets
  • Developing first Persian Automatic Question Answering systems on telephone for Irancell Customers

Skills

Machine Learning and Deep Learning Tools

  • Pytorch, TensorRT, ONNX, Keras and Sklearn, Tensorflow, Opencv, Pandas

    • Database Systems

  • MS SQL SERVER, SQLite, Redis, ElasticSearch, MongoDB

    • Operating System

  • Windows, Linux (Ubuntu)

    • Programming Languages

  • Python, Java, C/C++, NodeJS

    • Other

  • Redis, Docker, Git, RESTapi, Fastapi, kibana, matlab

    • Miscellaneous

  • LATEX, Photoshop, MS Office

    • Blog

    Deployment

    How to Convert a Pytorch Model into ONNX

    In this post you can read my medium story about converting pytorch to ONNX

    Fateme Bafghi
    5 min
    Deployment

    How to Convert an ONNX Model into TensorRT

    Here is my github repo on converting an ONNX model to TensorRT, Calibrating it and getting inference from it

    Fateme Bafghi
    5 min

    Languages

    Persian

    Native
    English
  • TOEFL Overall:106 Listening:30 Reading:26 Speaking:26 Writing:24
  • GRE Overall:313 Quantitative Reasoning:165 Verbal Reasoning:148 Writing:3.5

    • Researches

  • “Multiple-Vehicle Tracking in the Highway Using Appearance Model and Visual Object Tracking.”, Oral presentation The 11 Iranian and the first International Conference on Machine Vision and Image Processing (MVIP), (2020)
    • In recent decades, due to the groundbreaking improvements in machine vision, many daily tasks are performed by computers. One of these tasks is multiple-vehicle tracking, which is widely used in different areas such as video surveillance and traffic monitoring. This paper focuses on introducing an efficient novel approach with acceptable accuracy. This is achieved through an efficient appearance and motion model based on the features extracted from each object. For this purpose, two different approaches have been used to extract features, i.e. features extracted from a deep neural network, and traditional features. Then the results from these two approaches are compared with state-of-the-art trackers. The results are obtained by executing the methods on the UA-DETRACK benchmark. The first method led to 58.9% accuracy while the second method caused up to 15.9%. The proposed methods can still be improved by extracting more distinguishable features.