Another Change

It has been a while. The last few years at the Straw lab have been some of the most satisfying engineering of my life - building FlyMAD , wrangling many-camera real-time tracking rigs , and generally being paid to point lasers at flies in the name of science. I have learned an enormous amount, and I have very little to complain about (which, as long-standing readers will know, is itself a kind of achievement).

But there is a change afoot. I’m excited to say that, together with Max Hofbauer, I am co-founding a company. It is called Loopbio , it is based here in Vienna, and the pairing is a deliberate one: Max is a biologist (trained here at the University of Vienna, in behaviour and neurobiology), and I am the engineer. The whole point of us is that neither of us is sufficient on his own.

Our goal is simple to state and, I suspect, rather harder to deliver: to bring camera tracking, high-throughput video recording, image processing and laboratory automation to the natural sciences - to make this technology something a working scientist can simply use, rather than something each lab must build from scratch.

Which brings me to the why, because it is the honest part. Somewhere along the way I noticed something, and once noticed it was hard to un-notice. The technology we were building - real-time vision, closed-loop control, reproducible data capture - is genuinely transformative for biology. It is also genuinely hard to build. It wants an unreasonable mix of kernel and network tuning, machine vision, real-time control and plain software engineering, all sitting on top of the actual biology you were trying to do in the first place. So it ends up as a bespoke, one-off rig that works beautifully on exactly one bench, understood by exactly one person, and is more or less unrepeatable the moment that person leaves - then reinvented from scratch in the next lab down the corridor.

The scientists who most need these tools - biologists, neuroscientists, people who are emphatically not engineers and should not have to be - cannot easily get them. They spend their scarce, expensive time fighting infrastructure instead of doing science. I spent real effort making FlyMAD usable by our collaborators (GUIs, a single-file self-installing build, supporting exactly one Linux distribution so that I knew what I was supporting) and I still watched how much friction remained. That experience is the seed of all this. The apparatus should not be the hard part. The science should be the hard part.

So I am leaving the comfort of academia (mostly) to go and try. It is a leap, and I am under no illusions about the odds, nor that wanting biology to be easier is the same as making it so. There is, perhaps one day, room for the tools to start learning a little for themselves - though that is aspiration, not a feature, and I have learned to be wary of shiny new descriptors. But I would rather build the thing I kept wishing existed than wait for someone else to.

Apparently I am an entrepreneur now, which is a sentence I did not expect to write. More soon, once there is something to show. For now, wish us luck.