LinkedIn, Twitter, Substack, Medium, Youtube
I’m a 18 y/o tinkerer from Toronto, Canada who's obsessed with all things full stack, AI/ML, and blockchain dev. I enjoy building software tools to address problems large and small. I’ve previously built a Chrome extension for mental health support, an automated dispatcher and CAD management system to handle 911 calls, and a gamified smart grid design tool to optimize electricity production and distribution. In my free time, you’ll either find me trying to run cool git repos locally, beating my siblings in Monopoly, dancing or running!
Some of my history!
Previously, I've worked at RBC on the DevOps team, Youth Culture as a project manager, and for Brampton Focus as a developer. I also do AI consultancy work, guiding small and large businesses, and developing custom apps for their unique needs. During some of my contract work, I’ve automated leadership coaching with Leaderly.Inc, validated AI’s ability to help strengthen relationships by building an AI networking app with the Institute of Food Technologists and more. I’ve had the honour of presenting my personal projects on global stages at SXSW2023, C2Montreal, the Women of Influence Gala and more. I’m also privileged to have received support from Masayoshi Son, and Eric Schmidt by becoming selected as a Masason Foundation member and Rise Global Winner.
My journey in AI x Drug Design actually began in the summer after I fell down a rabbit hole of building reinforcement learning algorithms to play video games. In learning that this same tech was behind AlphaFold 2’s success in modelling protein structures, and that Insilico Medicine had used AI to develop an experimental drug for the incurable lung disease IPF, I knew I needed to work in this space.
I took courses to gain technical skills, read papers to understand the drug discovery process today, explored cutting-edge tools commonly used in drug discovery, replicated state-of-the-art models, and built a platform to speed up drug target identification and biomarker discovery with AI agents.
Reinforcement learning courses + YouTube tutorials
I’ve previously done loads of full stack development building web applications as well as more traditional machine learning work, but I was curious to learn about reinforcement learning so I took on the Hugging Face Deep Reinforcement Learning Course. Hugging Face has made it insanely simple to begin building with RL libraries like Stable Baselines3 and even showed how to use unique custom environments on top of the classics like Space Invaders. Reinforcement learning’s similarity to the way humans interact in the world made it interesting to learn the theory and math behind the algorithms, not just the code. Once I knew I wanted to work in drug discovery, I followed up by completing YouTube tutorials on building tools for computational biology.
Reading papers, exploring tools and replicating SOTA models
Next, I read lots of exciting papers exploring promising novel ML models. Insilico Medicine’s Generative Tensorial Reinforcement Learning (GENTRL) model, was one phenomenal example of RL in the field, but I quickly learned that there is so much more happening as well. Researchers have applied GANs, LLM’s and more. Each model having their own strengths and weaknesses. I explored as many ML applications in drug discovery as possible and wrote an article about how I wished drug discovery was super-powered. I even did an in-depth breakdown of one of the models that I found the most exciting the Regression Transformer.
Building a platform to automate data analysis, visualization and insight generation for biochemistry
Although I really enjoyed building ML models for drug discovery I found it challenging to work with the data. There was so much of it, and without a biological background, I felt disadvantaged in understanding hidden patterns in my data that could contribute to the development of better ML models. That's when I decided to build an agent assistant that could help me with my drug discovery tasks. I built BioBytes which is a platform to automate EDA through simple natural language commands helping researchers, scientists, ml develops and more make sense of the biological and chemical datasets they are working with quickly.
https://www.loom.com/share/171473f2ccf245578c3ec6c4cd8a5835?sid=90eb9766-77cd-469c-8282-c1f8d823d9ca
BioBytes currently serves as an educational tool for those aspiring to apply computational and data science techniques in their research, and can also act as a helpful automation tool for professionals in the field.
Currently, BioBytes can help you