DEVLog - April

April was a fantastic month, packed with new challenges and complex problems to solve. Let’s take a look at what I’ve accomplished.

Bezier

Weekend after weekend, I’ve been grinding on Bezier, my tool for building beautiful and accessible documentation. I’ve been sharing updates on X, but here’s a proper summary of the progress.

The first major milestone this month was merging the Drawmatic modules with the document processor into a single, cohesive app. The highlight? Real-time updates to titles and the work tree in the sidebar.

The second milestone was integrating the API with the client app. This was fairly straightforward since the current action set is still lean. With this update, “storytellers” (Bezier users) can now create stories—documentation pages—by selecting folders, documents, and diagrams.

Finally, the last big milestone was integrating the bucket service and enabling real-time collaboration. Now, multiple users can edit documents simultaneously without stepping on each other’s toes.

With all this progress, I’m confident I’ll be releasing a beta very soon. Let’s see what May brings.

AI and ML

As many of you know, I’m a professor of Artificial Intelligence and Machine Learning. This month, I implemented several projects to demonstrate the practical capabilities of the models we discuss in class.

ChromAI

The first project of the semester involved implementing an ML model to determine the optimal foreground color for a given background. It was the perfect example to introduce Logistic Regression.

ScreenBuddy

In the second class, we dove into recommendation systems and clustering problems. To bring it to life, we built a collaborative movie recommendation system using a KNeighbors model.

Identity

The third class focused on hyperdimensional problems, where I introduced SVM models. We used an SVC within a pipeline, combined with PCA, to build a face recognition application.

AlexNet

The final class of the month was an introduction to Deep Learning, featuring AlexNet, the pioneer of deep models. I guided the group through developing their first neural network. We implemented AlexNet from scratch to classify flowers, only to rediscover that training deep models from scratch takes… a while. We pivoted to using a pre-trained model (MobileNetV2), which crushed the task after just 5 epochs and 150 seconds of training.

The Next Steps

It was definitely an awesome month. I’m hoping May will be just as amazing, with major progress on Bezier and more fun, hands-on projects for my students.

---

If you enjoy the content, please don’t hesitate to subscribe and leave a comment! I would love to connect with you and hear your thoughts on the topics I cover.