About me

Hey there, looks like you found my blog and want to know more about me.

I am Moritz Wanzenböck, a developer and part-time student of mathematics. Among my favorite topics as a developer are writing high performance algorithms/data structures, operating systems and automation. As a mathematician I am also leaning close to IT, my personal favorites are cryptography and coding theory (as in compression algorithms and error correction).

I am currently located in Vienna, Austria. After having finished my bachelor's degree in mathematics I am on the lookout for a fitting master's programme.

Currently I'm working...

... for LINBIT as a software developer

Currently we are working on bringing the power of 20+ years of development on DRBD and LVM to the world of software defined storage in Kubernetes. Project include:

• Creating an open source operator to deploy our SDS solution
• Developing a k8s scheduler plugin to improve storage locality
• Improving our test set up by running automated tests against freshly provisioned k8s clusters on every merge
• Our in-house CI tools, notably virter, a faster alternative to vagrant.

... a pylint plugin that warns if you import something not specified in setup.py

At work we use a single repository to house a bunch of related python packages. Each package should be installable separately (some of them are "released" to the public, some are private) via pip. To make sure we don't accidentally miss specifying all (direct) imports, this plugin crawls all imports and verifies they are referenced in setup.py

... on a drop-in replacement for dvc run

dvc is a tool for creating reproducible pipelines, originally for data science, but also useful in other areas, such as machine learning.

One of the main commands when using ìt is dvc run, for example:

$ dvc run -f featurize.dvc \
          -d src/featurization.py -d data/prepared \
          -o data/features \
          python src/featurization.py \
                 data/prepared data/features

This describes a pipeline stage, which depends on src/featurization.py and data/prepared to produce data/features.

This is quite nice, but I think it can be improved. Notice that all inputs and outputs are specified twice, once for dvc to record, once for the actual command to run. Say for example your python code imports another python file. Then you need to also add this to the command line, but it is very easy to overlook.

My plan is to directly replace dvc run with an enhanced version that uses the ptrace API available on linux to record all input files and output files. It would then write a dvc compatible file to disk. This has the benefit that other commands should just work.™

Previously I worked...

... for cloudlight.io as a software developer

We were working on a cloud platform for the Earth Observation community, providing easy access to the Open Source GRASP framework. Some of the tasks include

• Writing python packages that can be used by scientist to access and analyze satellite data
• Servicing the cloud platform (mostly openshift)
• Writing high performance image processing code (mostly C++)
• Writing a low resource overhead task scheduler (Rust)

Older projects

• xch a library/CLI tool to swap paths atomically, written in Rust.
• rqrr a high performance QR code reader, written in Rust.
  • As a little diversion, I also implemented a rust macro to automatically structs that implement high performance finite field arithmetic over any binary field: g2p
• this blog mostly I just wanted to build a fast and responsive website with just CSS and html. Mostly to show myself that I could. If you notice any issue with this site, please reach out via the mail linked above.

Other interests

On most weekends you can find me somewhere in the Austrian Alps, either hiking or climbing up some mountain. There is a work-in-progress section of this blog dedicated to that here

I also enjoy playing board games, play the occasional video game and cooking.