MathWorks

It is the best of the companies to work for. It is the worst of the companies to work for. It all depends on where you end up in the company. MathWorks has two major product lines: one around MATLAB and one around Simulink. Both product lines are very successful and comprise the two halves of the revenue.

The engineering organization is built into two halves around these product lines and evolved very differently from each other. MATLAB has been a historically "Learn and Apply" tool where the market is composed of a very large number of small accounts. Simulink has been positioned as a serious engineering design tool which caters to large accounts in automotive, aerospace, etc.

I will compare and contrast these two halves, and it turns out that if you play your cards right, you can have a very nice life at MathWorks (just not a very successful one).

The MATLAB half is, in general, really fun to work for. There are not much market pressures as MATLAB users are mostly students and one-off engineers for whom MATLAB serves very adequately. The timelines for getting anything done in this organization are truly geological. People spend one year researching the requirements, the next year documenting the design, and the next in various design meetings and so on.

The management is organically grown, which means the hard-working years are well into their past, and they are allowed to coast and concentrate on finer things in life. The relaxed, hands-off management means there is even less pressure on the foot soldiers. The incredible focus on process (specs, design reviews, etc.) enables a large number of highly incompetent engineers to hide behind the process to get very little done. It is no accident that this organization is filled with native English speakers who are better at talking than in doing. There have been innovations on the MATLAB side, but they are rare, and they happened in spite of the system rather than because of it. The 90-10 rule applies very strongly here.

I do not have many good things to say about the Simulink side, so let us move on to the cons.

Working on the Simulink side is brutal, mind-numbing, and in general, career limiting as you will not be able to market these skills anywhere else. The management here is also organically grown, and many of them do suffer from the same malady as their MATLAB counterparts, where they rest on the laurels earned in the 1990s. There are a few exceptions, but they are just that: exceptions.

The upper management in this organization puts a lot of pressure on getting things done as fast as possible. Large projects are never allowed to even take off for the fear that they will never finish. The extreme risk aversion made this organization extremely incremental in its approach. There are teams that add no more than a few check-boxes to their product in a release cycle, but as long as they ship something, the management is happy.

There is very little support or recognition for those who work on infrastructure and silently contribute to the products. Because of this, no talented engineer wants to work on infrastructure for fear of being marginalized. This meant that those who want to work on infrastructure full-time are the lazy, talentless ones. This is causing this area to suffer a slow death which is certain to arrive in the next decade.

Force people to move between MATLAB and Simulink sides. This creates cross-pollination of cultures.

Get rid of managers who merely push paper around. A technical organization like this deserves hands-on managers who can lift the burden of the whole team if needed. This is especially the case in quality engineering, where incompetent managers hired even more incompetent engineers under them.

The so-called foundation services, such as usability and program management, are chock-full of inefficiency, as they really have very little to do. When you walk by their offices, you will see them trading stocks, playing on Facebook, watching YouTube, etc. They can all be replaced by hourly workers, and we will lose nothing.

MathWorks has become like Microsoft: big, bloated, ineffective, and unimaginative. It is probably because many of the processes at MathWorks are borrowed from Microsoft. We should find a way to become more like Apple or Google. It is easier said than done.

One idea is to set up small teams within the company and give them carte blanche on rewriting MATLAB and Simulink. Give them five years and a promise of a million-dollar bonus each if they end up shipping in five years.