C++ Developer / Computer Science (m/f/x)

Our focus is on business slides (as opposed to more artful applications) because they offer great potential for automation of layout tasks that are traditionally performed by PowerPoint users themselves. Challenges are plenty: from a solid understanding of what makes a good layout and which guidelines are followed by humans who do manual layout, to algorithms that produce an acceptable output fast enough for interactive slide design, to a graphical user interface that supports our new, original approach to slide layout in a way that is easy to understand yet unobtrusive, to solid technical solutions for automatic bug reporting and automatic updates, to compatibility with third-party software on the computers of half a million users.

Here are some highlights of what we have done.

Language

• Everything we do is C++. Even our customer portal is written in C++. There is some Assembler glue code where it is necessary, and our build scripts are written in Python, but other than that our company is all about C++.
• Naturally, we use C++ features like lambdas and rvalue references throughout our code base, and have switched to C++17 where our compilers support it.
• We fund the working group for programming languages of the German Institute for Standardization (DIN). Some of our employees are members of this committee and vote in the international standardization process of ISO/IEC C++.
• We sponsor the Standard C++ Foundation helping them to promote the understanding and use of modern Standard C++ on all compilers and platforms.

Library

• We use Boost throughout our code, e.g., Boost.Spirit for parsing.
• We have our own range library, in the same spirit as Boost.Range or Eric Niebler’s range-v3, but going further, for example, by unifying internal and external iteration. We gave a talk about it, and most of the code is public.
• We develop our own cross-platform library to support Mac and Windows with a single code base.
• We have our own reference-counting and persistence libraries to save and restore whole object trees.
• We have an extensive bug reporting infrastructure. Assertions and error checks stay in the release code, and our software automatically reports bugs to our server. The server analyzes the bug, categorizes it and files it in a database that all developers can access. If an update fixes the bug, the user can download the update directly from a bug response web page.

Algorithms

• our company was founded on the idea for an algorithm for automatic slide layout, and we are still on an exciting journey towards that ambitious vision. You can see our most recent release in action!
• We developed a new algorithm for automatic point cloud labeling that allows labels to be positioned away from the actual points.
• We developed a new algorithm for automatic column chart labeling.
• We are working with John Forrest – author of the linear solver CLP – to make his simplex code faster on our kind of problems.
• We developed many generic data structures that are not in C++ or Boost, for example partitions.
• Our software not only produces charts, it is also able read them back from paper. For our chart recognition tool, we rely on OpenCV and the Leptonica Image Processing Library.

Reverse Engineering

• We do lots of reverse engineering with the disassembler IDA from Hex-Rays, in order to achieve things that are not possible via the documented Microsoft Office API.
• We wrote probably the best function hooking engine out there. On each start of our software, we patch the Microsoft Office executables in memory. Rather than hard-coding patch addresses, we search for small chunks of assembly code to be robust against minor changes in the executables.