DEVLog - April

April was a great month, full of new challenges and problems that I had to solve. Let's review what I have done this month.

Bezier

Weekend after weekend I have been working on Bezier. A tool to build beautiful and accessible documentation. I have commented my progress on X, but let's make a summary of my work.

The first milestone reached this month was merging the modules of drawmatic and the document processor I was implementing in a single app. The most interesting thing was the real time changes on the titles and work tree at the sidebar.

My second milestone was integrating the API with the app client. It was easy as there are few actions and resources at the moment. Since this milestone, "storytellers" - users of Bezier- are able to create stories - documentation pages - selecting between folders, documents, and diagrams.

Finally, my last milestone of this month was adding the integration with the bucket service and managing the real time collaboration, so users can edit documents at the same time.

A lot of progress during this month make me feel confident I will release a beta pretty soon, let's see how it goes in May.

AI and ML

As many of you already know, I'm a professor of Artificial Intelligence and Machine Learning and this month I implemented many projects to teach my students the capabilities of some of the models we talk about.

ChromAI

The first project of the semester was implementing a ML model to determine the foreground color given a background color. I used this example to introduce the Logistic Regression and its capabilities.

ScreenBuddy

The second class I talked about recommendation systems and clusterization problems. For this, we developed a collaborative movie recommendation system using a KNeighbors model.

Identity

The third class was about hyperdimensional problems and I presented to my students the SVMs models, using a SVC in a pipeline with a PCA to create an application to recognize faces.

AlexNet

The last class of the month, my lecture was an introduction to DeepLearning, presenting the first model that can be considered deep, AlexNet. Here I guided my group to develop their first neural network and then we implemented AlexNet by its layers diagram to classify flowers, just to discover that training our own models sucks, so we decided to use an existing model (MobileNetV2) that outperformed the task after 5 epochs and 150 seconds of training.

The Next Steps

This was definitely an awesome month, and I really hope this new month will be amazing with a great progress with Bezier and funny and interesting 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. Your feedback is greatly appreciated! Leave a tip.