Verified Expert in Engineering
Machine Learning Engineer and Developer
Reza is a highly skilled machine learning engineer with eight years of experience. Currently working at IBM, he has developed and deployed cutting-edge machine learning algorithms to enhance virtual assistant products. His research at the University of Toronto focused on numerical modeling techniques for solving multiphase flow problems. With a strong background as a machine learning engineer and data scientist, Reza's expertise lies in natural language processing and computer vision.
MacOS, Slack, Visual Studio Code (VS Code), Linux
The most amazing...
...thing I've deployed is a transformer-based model for paraphrasing irregular user utterances, significantly improving the quality of responses to user questions.
Machine Learning Engineer
- Developed and deployed machine learning solutions using large language models to enhance text classification and automate model improvement based on user feedback with limited data, resulting in six patents.
- Created and implemented clustering algorithms to unify unrecognized utterances into intents for model training.
- Built a unified metrics framework to evaluate improvements in machine learning algorithms for text classification, resulting in a published journal article.
Machine Learning Engineer
- Led a team of four in developing sequence-to-sequence models using PyTorch to rewrite irregular questions to well-structured ones, helping our natural language understanding (NLU) layer understand poorly structured questions.
- Guided a team of three in creating phrase suggestion models given a list of keywords using pre-trained transformers-based language models.
- Developed a neural retrieval system coupled with a sequence-to-sequence model to respond to factual queries.
- Led a team of three data scientists and data engineers to build Looka's recommendation and smart suggestion engines using Elasticsearch, increasing the offline conversion rate, specifically user sign-up to package purchase, from 5% to 9%.
- Trained deep generative models like convolutional VAEs and GANs in PyTorch and RNNs in TensorFlow to generate design assets, allowing users to access high-quality, royalty-free fonts and symbols.
- Supervised two data science interns over four months, helping generate over 1,000 design assets such as fonts, symbols, and layouts using deep generative models developed in PyTorch and TensorFlow and trained on AWS.
- Collaborated with a team of four data scientists to create an in-house development framework for building pattern recognition algorithms aimed at a wide range of biometric applications based on the electrocardiogram (ECG).
- Developed pattern recognition algorithms using techniques such as principal component analysis, linear discriminant analysis, and autoencoders for feature learning, k-nearest neighbors, support vector machines, and neural networks for classification.
- Trained a convolutional variational autoencoder using TensorFlow and optimized it for maximum performance and minimal memory footprint for deployment on ARM Cortex-M CPUs targeted for intelligent IoT edge devices.
Unity Health Toronto
- Developed SVM and MLP models for classifying neonate breathing patterns based on features extracted from the electrical activity of the diaphragm, offering an unbiased approach for categorizing neonate breathing.
- Implemented signal processing algorithms in Python and C for real-time signal analysis on the Arduino platform.
- Communicated analytical insights to industry partners using interactive dashboards developed in Excel VBA, securing more than $100,000 in research funding.
University of Toronto
- Developed an open-source numerical modeling software (OpenPNM) in an international collaboration between four academic institutions, providing over 100 scientists worldwide with a framework for performing multiphase simulations.
- Developed, tested, and executed parallelized C++ codes on SciNet, a high-performance computing network running on Unix, allowing for the analysis of large datasets within the 48-hour job execution time limit of SciNet clusters.
- Performed unit tests using the Python unit testing framework, unittest, and prepared code documentation using Sphinx.
Automated Dataset Pairing Algorithm
Automated Virtual Assistant Learning
Automated Font and Symbol Generation
Python, C, Excel VBA, SQL, C++
PyTorch, TensorFlow, OpenMP
Slack, Git, GitHub, AWS CLI, MATLAB
Data Science, Parallel Programming
MacOS, Visual Studio Code (VS Code), Docker, Kubernetes, Linux, Arduino
Machine Learning, Language Models, Natural Language Processing (NLP), Computer Vision, Recurrent Neural Networks (RNNs), Convolutional Neural Networks, Artificial Intelligence (AI), APIs, Large Language Models (LLMs), Generative Pre-trained Transformer 3 (GPT-3), AI Programming, Deep Learning, Algorithms, Analytics, Big Data, Unstructured Data Analysis, CSV File Processing, Generative Pre-trained Transformers (GPT), Chatbot, Data Analytics, Data Analysis, Mathematics, GPT, Chatbots, Chatbot Conversation Design, OpenAI, Technical Leadership, Mentorship & Coaching, Deep Neural Networks, Image Generation, Generative Adversarial Networks (GANs), Variational Autoencoders, Signal Processing, Digital Signal Processing, Open Source, Numerical Modeling, API Integration, Leadership
Master's Degree in Engineering
University of Toronto - Toronto, Ontario, Canada
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
Share your needs
Choose your talent
Start your risk-free talent trial
Top talent is in high demand.Start hiring