I've got a team of software developers, machine learning people, and scientists, and we're developing a range of methods for solving different data analysis problems, most of them around sonar and underwater acoustics. We've also got some virtual reality displays in the mix for interpreting 3D underwater acoustic data and for training sailors to safely command helicopters on ship decks. There is also a project modelling the behaviour of the flow of air around ships to help simulate the safe operating conditions for landing a helicopter. There's all sorts, but it's all mainly around physics modelling and signal processing to help solve various problems.
I started out in aeroacoustics, where I did a PhD making wind tunnel measurements of wind turbine noise using an acoustic array. The acoustic processing is very similar to how a sonar system determines where a sound is coming from. Post PhD I went into engineering consultancy where I worked on a variety of mechanical and aeronautical engineering projects, but I continued to work on several acoustic projects, including designing an acoustic array for making water tunnel measurements of flow noise from ships.
After that I worked for a company based up in Edinburgh who manufacture very high-frequency 3D imaging sonar systems, where I developed novel beamforming methods, getting my name onto a fair few patents. When I moved back down to the southwest of England, I was looking for a job in sonar in this area, and had worked with SEA on previous projects, so knew it was a decent company.
When the previous technical lead for a project I was working on left SEA, I agreed to take it on, and my role has just grown from there. Over the last three years, I've picked up more and more tasks and now have a team of developers working on all the projects described above.
There's a lot of scope for making a difference and changing things, and that's partly because we're inevitably charting new territory in Research and Simulation and partly because the company is growing fast, and the way we work is evolving to support this. Sometimes this can feel challenging, because you're often doing things for the first time, and there's not the support of “we've done this year's for years, and this is the process, and this is how you do it”. The flip side is that you get the chance to shape that process and build a new team from the ground up.
A lot of the output from our simulation work is very complex, and this can make it difficult to prove that it is “correct” and to demonstrate progress. For example, if we’re simulating acoustic propagation from a battleship to a submarine sonar array, the output is a big pile of numbers. If we were modelling the propagation of light rather than sound it would be much easier, because I know what a battleship looks like, but do I know what it sounds like from 10,000 yards away on a windy day in the North Atlantic?
This is where we have to rely on our ASW experts’ experience of actually sitting in a sonar operations room to put on a pair of headphones and tell us whether it sounds right. To get all of it working, we have many moving parts: there's the underlying physics modelling, the software development, deployment to a large server computer, networking, interfaces to other software on both the inputs and outputs, the list goes on. Getting these components working together is a real challenge and requires a team of experts from different areas of the business to bring it all together.
As my team gets bigger, I get to do less and less actual coding, which is a bit sad, as I do love getting stuck into some proper software development. I think everyone in my team would agree that it’s underrated as a creative activity. When you get on a roll with it, it's like painting or playing music or writing, because it takes all of your focus, and the time just flies past.
One thing we're doing quite a bit with is GPU processing. Rather than running things on the normal chip on your computer, you run it on the graphics chip instead, and there's various computational benefits to doing this. We’re using this approach for our latest developments in sonar signal processing, and it underpins most of the recent developments in AI technology.
In my PhD lab we had PlayStations because they had the best graphics cards at the time, and people were learning how to write fluid dynamics calculations that were running in place of the calculations that would normally be working out what colour to make each pixel on your screen. The benefits of this GPU processing are now widely recognised, and there is an entire market of making graphics cards that are specifically for numerical analysis rather than for actual graphics. The vast majority of the computing behind things like ChatGPT are run on specialised GPU processors.
Don’t be afraid of the maths! When I first started out in acoustics, I wasn’t necessarily great at it, but my PhD supervisor encouraged me to acknowledge it’s hard and apply myself to it. Nobody really has an intuitive understanding of something like this, so you just have to accept that it's going to take a while for you to build your understanding. Once you have that understanding however, it will set you up for being able to be able to build organised and efficient simulations. One of the key things in simulation is knowing what you need to include and which things you can exclude or approximate. You can only make these decisions if you have an understanding of the underlying physics and how it is mathematically represented.
Thank you, Charlie.
