Hi, I’m Isaac.
Builder of data, ML and AI systems.
I do applied AI while keeping product, UX, effectiveness and production environments at the forefront.
I define evals and success criteria, then experiment.
Love good challenges.
Selected work
David
I plan my week in Telegram. If David suggests moving a meeting, I get the final say.
David reads my Google Calendar and uses summaries its synthesisers write after our planning sessions. It also draws on my long-term goals and the background I’ve shared over time. I approve calendar changes in Telegram before David applies them. David runs on AWS Lightsail.
AdReady AI
I led engineering on a tool for reviewing video ads before launch. We split media processing from inference to keep reviews responsive.
At PM Accelerator, I wanted brands to be able to review ads without slow media work holding up inference. We ran edge inference and media processing as separate services. That gave us two services to operate, but we could manage each workload separately. We handed off the MVP.
Snowpea
Book with nail and lash artists who work from home.
I worked on bookings and payments so customers could book with independent artists and get a clear confirmation. Snowpea used Stripe for payments.
ML Model Monitoring
The model shipped. How do I know it still works?
I built a FastAPI service and a Streamlit dashboard to track what happens after deployment. I use a KS test to flag changes in the input data without waiting for labels. I track accuracy and F1 separately because drift alone can’t tell me whether predictions got worse.
Cocoa Futures Price Modelling
Three ways to forecast cocoa prices. Which one held up?
I backtested ARIMA-GARCH, ETS and VARX on historical cocoa futures prices. VARX held up best in the comparison. The scripts let me rerun the evaluation when I changed a model.
Transformers and Autoencoders
What can a model learn from the parts of an image it can’t see?
I wrote this to understand how transformers and autoencoders fit together. I started with masked image reconstruction and followed the connections to visual classification and teacher–student training.
About me
I studied math, stats, economics and finance at University of Toronto.
I’ve done data work and analytics engineering at Dayforce and The BRIDGE, and AI engineering at PM Accelerator (AdReady AI & YourVillage).
If you’re putting agents & AI into everything while ignoring engineering + data principles, just to call yourself “AI-enabled” and “transformed”, I am NOT your guy.