Roughly follow a few selective chapters from the book 'Hands-On Machine Learning with Scikit-Learn and TensorFlow'

Currently, As QA, we design matrix for the test cases, then use matrix to generate test cases' name, then according the test cases' name to add related settings, then add those cases' name and settings to the openQA test suite. I hope, we can design matrix first, then other steps will be generated by automatically.

Metal³ is the idea to support a declarative bare metal cluster management for Kubernates by employing a simplified stand alone version of Ironic.

Currently, way we distribute ssh keys within QAM on our testing infrustructure is a bit cumbersome - maybe we should try to (ab)use existing salt used by our internal infra team.

We kinda need more flexible PXE in Prague office, UEFI would also not hurt - so lets check what we can do to make it better.

Pi Security

In the past, some customers ever complained that OCFS2 file system performance went down when listing a big directory (e.g. include 400k files). According to my preliminary investigation, this performance problem is related to file system block allocation algorithm. Then, I want to look at the current mainstream file systems (e.g. EXT4, XFS, etc) block allocation algorithms in this hack week, compare and analysis these algorithms advantages/disadvantages.

RISC-V is an open ISA (Instruction Set Architecture) based on RISC architecture. It's originated from UC Berkeley and it's attracting more attention in recent years because of its full open architecture so every developer has opportunities to get involved in application processor design or apply it into different applications, such as IoT, Robotics, ... etc.

I just recently started a new project which is meant to be a wrapper for OBS's API. The library is written in Ruby.