Hi, I'm Gurkamal Deol

I'm a Machine Learning Engineer and Biologist. My interests include AI driven drug discovery, building end-to-end ML apps, & writing about science.

Free Template

Toxic SMILES

Visit web app

Toxic SMILES

Full stack machine learning web app that predicts whether a molecular compound is an environmental carcinogen. Built using RDKit, Flask, Docker, Microsoft Azure, Scikit-learn, and Bootstrap. Read in-depth about it here.

Free Template

Frodotype

Visit web app

Frodotype

An end-to-end machine learning web app that generates fantasy themed text. Built using the GPT2 model, Tensorflow, Chart.js, Google Cloud Platform, Docker, and Bulma CSS.

Free Template

Drug Decoder

Visit web app

Drug Decoder

My very first web app which I used for converting lists of drug IDs to SMILES strings. Built using Flask, Pandas, Numpy, and Google Cloud Platform

Free Template

Federated Learning

Read on Medium

Living & Computing on the Edge - Federated Learning

A Brief Introductory article on Federated Learning.

Free Template

Maximum Likelihood Estimation

Read on Medium

Maximum Likelihood Estimation

An article going through the math behind calculating Maximum Likelihood Estimation by hand.

Education

  • Masters — 2019

    Bioinformatics

    The University of Guelph

    Masters Research Project — Exploring the Feasibility of Novel Small Molecule Patentability Through Building a PostgreSQL Database and Identifying Tanimoto Nearest Neighbours.

  • Bachelors — 2017

    Honours Specialization in Genetics; Major in Pharmacology

    The University of Western Ontario

    Honours Thesis — Detecting Prostate Cancer Using Tumor-Activatable Minicircles Encoding the Secretable Biomarker SEAP

Work Experience

  • Data Analyst

    Loblaw Companies Limited - Present

    Automating daily tasks, coverting SAS projects into python scripts and performing data integrity checks on enterprise data.

  • Bioinformatician

    Cyclica - 2019

    Built a cheminformatics data pipeline, put together a PostgreSQL database of 17 million small drug molecules, and integrated KNN/Tanimoto search functionality to the database.

  • Data Science Fellow

    SharpestMinds - 2019

    Built a full-stack machine learning application to predict environmental carcinogens, converted molecules to binary fingerprints and trained logistic regression, knn and gradient boosting models - achieving a final accuracy of 96.4%.