ML / MLOps experiments
Building small ML systems and learning the operational side of them.
I like exploring how models are trained, served, and maintained in the real world. The fun part is not the notebook alone — it is making something usable, observable, and repeatable.
Model loop
Draw a simple sketch and send it to my QuickDraw classifier. The model currently predicts bicycle, cat, and coffee cup.
The browser converts the drawing to a 28×28 grayscale matrix before it is posted to the serving endpoint.
White strokes on a dark background best match the training data.
Why this area excites me
Experimentation
Small ideas quickly become testable systems with real feedback loops.
Delivery
The value is in making the model easy to run, observe, and maintain.
Learning
This is where software engineering, data, and product thinking converge.