Mohamed Harmanani Email & Phone Number

Toronto, Ontario, Canada

Kingston, Ontario, Canada

Kingston, Ontario, Canada

Kingston, Ontario, Canada

Montreal, Quebec, Canada

◦ Accelerated the development of MLOps tools using DVC & Google Dataflow ETL to build data processing pipelines for NLP◦ Optimized deep learning models (BERT) for financial language modelling using PyTorch and AWS SageMaker, resulting in a 33% increase in performance metrics browsers, increasing volume of collected data by 20%◦ Built enterprise level data visualization tools for data exploration by product teams using Streamlit, providing real-time visibility into relevant KPIs◦ Utilized explainability frameworks to interpret Machine Learning models, turning predictions into product improvements

Toronto, Ontario, Canada

• Implemented NLP and deep learning techniques (RNN, BERT) for automated bug detection in student code using PyTorch• Improved data processing tools for extracting 15000+ data points from student submissions to serve as training data• Trained and fine-tuned RNN, Transformer, and CodeBERT models, leading to a 55% increase in model performance• Implemented a front-end tool for generating 2500+ test inputs on the fly, decreasing time spent testing models by 300%• Deployed over 50 machine learning models to a Flask web server, allowing to run and compare models 5x faster

Mississauga, Ontario, Canada

• Led the development of a data pipeline for biological sequence analysis using Slurm, Bash and Python, reducing task processing times from 1 week to 2 days• Developed tools to clean, assemble, and align RNA transcript data, using FastQC, Trinity, Hisat2, and BLAST• Wrangled over 6TB of data, and extracted insights such as alignment frequency and average transcript length

Toronto, Ontario, Canada

• Developed an AI-assisted design tool for infographic generation using React and Fabric.js, allowing users with no design experience to generate compelling infographics• Implemented probabilistic algorithms in TypeScript, increasing quality metrics of generated infographics by 13%• Developed a scalable service for icon searching using React & Algolia, reducing search times on the icon library by 200%• Identified pain points and performance issues by logging and analyzing user data using Mixpanel, Python, JavaScript• Helped over 250 users by resolving various technical issues reported to customer support via Intercom live chat

Doctor Of Philosophy (Phd), Artificial Intelligence

Master Of Science (Msc), Artificial Intelligence

Honours Bachelor Of Science (Hbsc), Computer Science

French Baccalaureate, Scientific Track