I finally had the time to finish this huge update, what I started developing more than a year ago: on-device machine learning. Users now can analyze device sensor data with an on-device machine learning algorithm to predict driver aggression.

The ML model can be downloaded via Firebase ML service. The development of the model is also fully open-sourced and available under the Hawk AI GitHub repository with scripts to train and convert the model to the TensorFlow Lite format.

Plans​

• Rework session screen with better data visualization.

For the foreseeable future, I don't plan to add more core features to the app, but when something comes up, I will resolve any issues.

I hope this will change in the future, but for now, enjoy what I consider a solid open-source base for Android application development.

This site was created to document changes and have a landing page of the Android version of Jay.

I plan to experiment with blog posts, like this, to get the hang of managing this side of a production ready-ish app. A brief description of Jay, and what the project is aiming for:

Jay is a data analytics solution, which aims to make data analytics more accessible for Everyone. The Android app collects data on-device, which can be uploaded to the cloud to be analyzed and make local decisions on local problems.

Features​

• Collect and display sensory and location data
• Upload and sync data in the cloud
• Support multiple users on a single device
• Provide an easy-to-use and good looking interface
• Be cheap to implement and maintain
• Analyze data

Plans​

• digitally generate and label data, which then can be used to train machine learning models
• analyze/process data on-device with ML models
• enable users to have supervisors for B2B applications
• make an administration app/webpage to supervise users
• make even more tools and solutions accessible, be as transparent as possible during development