Reza Torabi

Senior Data Scientist

2023 - Present

Baker Hughes

• Developing an attention-based classifier with vision transformers (ViT) to recognize defects, cracks and metal loses in oil and gas pipelines as well as deploying it in the company’s desktop application in the production phase.
• Image to image translation of 2D sensory data to 2D laser dig data via image segmentation techniques and UNet to estimate the size and depth of defects in the pipeline.
• Created a growth rate prediction model using deep neural networks to predict growth rate of defects in oil and gas pipelines.
• Developed a classifier to discriminate normal, disturbed, and reverberated sensory data and deploying it in the related software in the production phase.
• Developed a sensor failure recognition model.
• Developed an effort prediction application to predict, visualize and analyze effort.

Data Scientist

2021 - 2023

Aurora Imaging Group/University of Calgary

• Created an image retrieval system and image similarity search ranking engine for a web application using deep neural networks (noisy Autoencoder) that recommend space science community members similar phenomena and events in large Aurora image database according to their similarity rank.
• Developed Aurora image classification models for a web application using deep neural networks and transfer learning that generalized well on large Aurora image database with stat of the art accuracy. Cloud computing was used in training phase and Deep Convolutional Generative Adversarial Networks (DCGAN) for data augmentation to enhance the accuracy.
• Developed an image labeler and viewer desktop application for viewing, filtering, searching, and labeling (for supervised learning) the Aurora images.

Machine Learning Researcher

2019 - 2021

Canadian Center for Behavioral Neuroscience

• Developed a deep learning based behavioral recognizer toolbox for analyzing behavioral video data using CNN and RNN (LSTM) and knowledge extraction tools such as DeepExplain.
• Predicting cognitive score in children using pose estimation methods based on analyzing videos of their Lego building.
• Cage monitoring (detecting and tracking animals in their cages) using object detection and tracking methods using computer vision.
• Developing machine learning algorithms based on how brain works such as recirculation, Hebbian, and backtracking algorithms from the scratch.

Computational physicist

2018 - 2019

Complexity Science Group/University of Calgary

• Simulation, computational and statistical analysis of pattern formation.

Lecturer & Researcher

2010 - 2018

Tafresh University

• Teaching statistics and researching on statistical physics and network science.

PhD in Physics

Amirkabir University of Technology

MSc in Statistical Physics

Amirkabir University of Technology

BSc in Physics

Sharif University of Technology

Programing language

Python, C++, C#, Java, JavaScript, Bash, SQL, Swift, Matlab

Machine learning algorithms

Linear and Logistic regression, Decision tree, Random Forest, SVM, Naive Bayes, K-means, hierarchical and density-based clustering, Genetic algorithm, PCA, LightGBM, XGBoost, Matrix factorization, Collaborative filtering, Zero-shot classification

Deep learning

Tensorflow, Keras, PyTorch, PyTorch Lightning, Deep Neural Networks, Convolutional Neural Network (CNN), Recurrent Neural Networks (RNN, LSTM, GRU), Transfer learning, Attention Based CNN, Behavioral Cloning (Self-driving cars), graph learning, Class Activation Maps

Computer vision

Image processing, Image recognition, Object detection, Object tracking, Optical flow, Pose estimation methods, Semantic and instance segmentation, Action Recognition

Generative AI

Autoencoders, Variational Autoencoders (VAE), Generative Adversarial Networks (GAN), Transformers, Diffusion models, Vision Transformers, Multimodal models

Reinforcement learning

Q-learning, Deep Q-learning (DQN), Actor Critic Method, Deep Deterministic Policy Gradient (DDPG)

NLP & Agentic AI

Sentiment Analysis, NER, Topic modeling, Question-answering, Translation, Transformers (BERT, GPTs), LangChain

Big data and Cloud computing

AWS, Sagemaker, Microsoft Azure, Google Cloud, PySpark

Development

Git version control, Docker, Django, Flask, FastAPI, Tkinter, PyQt5, Full stack web development, IOS development

Robotics

Raspberry Pi, ROS2

Databases

MySQL, MS SQL Server, Oracle, MongoDB

Data Visualization

Plotly, Dash

Statistics & mathematics

Bayesian statistics, Linear algebra, Probability theory, Monte Carlo method, Network science

Other skills

Recommender systems, learning to rank (LTR), Signal processing, time series analysis, TensorBoard, AutoML, MLflow, AutoGluon, linux, Gradio, Hugging face, Pygame, OpenAI API

Machine learning applied to data from the THEMIS All-Sky Imager Array: Clouds and APA, AGU conference 2021, New Orleans, USA (2021).

A Neural Network Reveals Motoric Effects of Maternal Preconception Exposure to Nicotine on Rat Pup Behavior: A New Approach for Movement Disorders Diagnosis, Frontiers in Neuroscience (2021).