One Zero One: Peter Bebbington, Machine Learning and AI Solutions
In our most recent episode of the One Zero One Podcast, we had a compelling discussion with Peter Bebbington, CTO & CDO at Brainpool.ai around Machine Learning and AI Solutions.
Brainpool.ai operates as a global network of over 400 Artificial Intelligence and Machine Learning experts who help software companies enhance their products with Machine Learning & AI.
Brainpool’s mission is to democratise access to AI and bridge the gap between academia and the corporate world by finding applications of their data scientists’ research in business.
How Brainpool Came About
Brainpool.ai, founded in 2016, was funded by the Computer Science Department within UCL (University College London) which is where Peter undertook his PhD in Physics. At UCL, Peter was exposed to many talented researchers working on the frontiers of Machine Learning & Artificial Intelligence. He and his co-founders saw that there was a demand for this specialist knowledge outside of academia in the commercial world, and the idea for Brainpool.ai was formed.
Brainpool operates as a consultancy, they work with various different Finance & Fintech clients and their technology gets used predominantly to make financial forecasts. During his interview on the One Zero One podcast, Peter explains how on the other side of the business, they are more product focussed, using the Brainpool platform to help figure out how to productise this technology.
They (Brainpool) use the consultancy platform to identify what should be produced rather than just creating a product and hoping there will be a market for it. Brainpool looks at the needs and wants of their clients who may be interested in particular API’s instead of building things in-house, and they then build products with that in mind. Brainpool has developed two products: Forstack, which predicts financial time series models and DAISY, the first fully automated construction design system powered by Artificial Intelligence.
Machine Learning & Artificial Intelligence
During the interview, Peter goes on to highlight that one of the most common difficulties for organisations at the moment who are entering or trying to use Machine Learning is the deployment aspect because models such as Forstack and DAISY are not static; they require constant re-training and re-deployment. Therefore companies need to have very complicated infrastructures in place to deploy and manage these models in production. In Peter’s opinion, this technology is only starting to catch up now to how people want to use this in production.
DAISY
As mentioned above, DAISY AI is Brainpool’ s recent venture in the construction space. It emerged from a partnership with a large UK based construction company who specialise in building timber flooring and staircases. Peter explains in detail the current floor planning process and how this requires a large level of human interaction, time and energy. He states that many elements within those designs include quite tedious and repetitive tasks which, in truth, a highly trained architect should not be doing. He explains that if a client wants to double their production, they would have to double their workforce, which as Peter notes, is not a scalable solution and this is where DAISY comes in.
DAISY automates those tasks and it does it in a way which is comparable in terms of costing — in fact sometimes it actually comes out cheaper than what a human architect would design. It also checks heuristics like integrity and any sort of regulations that the floor design has to abide by. Overall Peter explains how this model makes peoples lives in construction easier and more productive.
“We’re not trying to replace anyone, we’re just trying to make worker’s lives in the construction industry easier and more productive. We want to take all of the tasks that would take an architect a long time to do, and automate them.”
Peter Bebbington, Brainpool.ai
How Brainpool’s Experts Work With Technology
As stated by Peter, how Brainpool’s experts work with technology can be very mixed depending on legacy systems that many companies work with, especially financial industries, but there does seem to be more of a focus towards Microservices currently. The reason that organisations are favouring microservices is that they are seen to be much easier to manage in an operational and resource sense rather than the monolithic approach of the last decade.
When it comes to prototyping models, a very popular tool used is Python, but for more production focussed models C++ is very popular. However, in Peter’s opinion the technologies that are used are not the main priority, what’s more important is how the software is delivered and deployed. Any software that Brainpool build, they abide by the protocols of the company they are contracted by to create that software. They tend to containerise a lot of their software for their clients to be able to integrate it into their own infrastructure, whether this is On-Premise or in the Cloud.
Tips for Companies Looking to Attempt Machine Learning & AI Projects In-House
Peter advises that it is essential for companies to upskill their current workforce to lay the foundation to adopting AI and Machine Learning. Upskilling not only adds value for the business, but it can also be very motivating for somebody who perhaps does not have a very technical background to be able to learn new skills and technologies. He explains how Machine Learning technologies move so quickly and how many companies learn one implementation method, then a couple of weeks later a new one comes in. Peter stresses how companies need to adapt to a continuous learning approach in order to achieve success in Machine Learning and AI.
You can listen to the full podcast here where Peter gives more detail around each of the areas I’ve explored in this blog post.
You can find out more, stream and download any of Version 1’s One Zero One Podcast episodes from our website or any of your favourite streaming platforms.
