Pierre Baqué of Neural Concept tells us how they provide creative minds and computers the simplicity of software to invent, design and engineer together.
First of all, how are you and your family doing in these COVID-19 times?
Pierre Baqué: We are doing fine. Since we were already doing partial home office before COVID19, it was easy to switch to 100% home office, as it was deemed obligatory by the Swiss authorities. Some of our employees took advantage of the situation by working from home in their home country/region, close to their family. Fortunately, despite numerous tests done by me, or our employees, we didn’t have a COVID-19 case yet.
Tell us about you, your career, how you founded Neural Concept.
Pierre Baqué: I am currently the CEO and co-founder of Neural Concept. I received an engineering degree in Applied Mathematics and a Master’s degree in Operations Research from Ecole Polytechnique in 2013. After working as an Engineer for Credit Suisse in London, I joined the Computer Vision Laboratory at Ecole Polytechnique Fédérale de Lausanne. I finished my doctoral degree in early 2018 under the supervision of Prof. Pascual Fua and Francois Fleuret. My research focused on Deep Structured Learning, and Variational Inference applied to Computer Vision. As a result of my doctoral thesis, I took the leap as an entrepreneur and founded Neural Concept in 2018, through which I acquired important business and commercial knowledge.
The history behind the founding of the Neural Concept is as follow:
Around 2012, the quick progress of Convolutional Neural Networks in Computer Vision put an end to the so-called “Winter of AI”. Boosted by new mathematical techniques, by GPU computing and by the availability of data, Computer Vision researchers were able to solve tasks that were beyond reach a few years before. It is around the same period, in 2013, that two researchers, Timur Bagautdinov and I, started our PhD in one of the most renowned computer vision laboratories in the world: the CVLab of EPFL led by Prof. Pascal Fua. Instead of applying DeepLearning techniques to image processing, as was the trend in research at the time, we, together with PhD advisor Francois Fleuret, decided to explore the application of such techniques to graphical modeling, meshes and 3D geometric data. From there to processing CAD models and simulations, only one step was missing, and the connection was made during a personal vacation with a close friend of mine, Clément Nardari, simulation engineer for Dassault Systèmes. Soon after, CVLab became one of the first labs to work on mixing Deep-Learning with simulation and CAD data in 2016. A small group of researchers was created to work on these topics at EPFL, with Master students, PhDs and PostDocs, supported by technology transfer investments from EPFL. As a group, we pushed further the theoretical capabilities of the approach and started developing a software library called NCS “Neural Concept Shape”.
In May 2018, Neural Concept was founded with the goal of commercializing this technology. At the time of the foundation, several industrials had already expressed their strong interest in technology, and this is what motivated the creation of a commercial entity. Most of the members of the team from EPFL joined the startup. Between May and December 2018, Neural Concept’s team completed four pilot projects with major actors of the Aerospace and Automotive industries. All of these pilot projects were successful, which set the company on the track of growth.
How does Neural Concept innovate?
Pierre Baqué: The manufacturing industry faces unprecedented challenges. Major technological transformations need to happen under time and cost constraints, which sets pressure on the whole sector.
In this context, the following targets a clearly set within large engineering-driven organizations:
o Reducing time to market
▪ Reducing the Request For Quotation process time, especially in the context of companies acting as suppliers for a specific type of parts.
▪ Achieving better and faster-customized designs.
o Unlocking new engineering methods and much faster design iterations.
▪ Better product in less time as a result (concrete effects like weight reduction, higher durability, etc.)
▪ Better use available data and create more with them.
▪ Act as a complement and a magnifier to existing engineering tools.
To support this revolution and obtain outstanding engineering performances in reduced time, numerical simulations tend to become a guide to design rather than a mere final validation tool. However, it turns out that Computer-Aided Engineering (CAE) and optimization are still not in the hands of each engineer in such corporations. The main reasons are slow reaction times, price and required skills.
Fortunately, Neural Concept Shape is coming to help. NCS is unlocking the possibilities of CAE, already today making high-fidelity simulations accessible to a vast audience of engineers within some top organizations.
Interactive engineering simulation and optimization are made possible by untapping the unreasonable power of data. Predictive models that generate post-processed physical simulation results, using a raw CAD model as sole input, are now available. And they can learn from the data that your engineers anyway produce every day. These predictive models are being used to simplify processes and emulate simulation engineers’ expertise by transferring it to design engineers early in the development process. It makes it possible to reduce the number of iterations between teams, optimize and accelerate design activities.
Neural Concept Shape is a software platform that lets engineering teams and companies implement these new workflows efficiently by exploiting the different levels of skills between simulation, methods and design teams.
How the coronavirus pandemic affects your business, and how are you coping?
Pierre Baqué: The productivity has not been impacted by the pandemic. However, social interactions between employees have been reduced due to constant home office, which decreased the overall “team spirit” of the company. However, we try to organize events, whether via videoconferencing or face-to-face, to maintain and improve relationships between the employees. On the business side, the company has been impacted due to the impact of the pandemic on some customers, especially on the Aerospace sector, but the business is regaining traction since the beginning of 2021.
Did you have to make difficult choices and what are the lessons learned?
Pierre Baqué: We had to prioritize our strategic directions and adapt our funding plans. We learned that anything can happen and that we need to be ready for tough times at any time. What specific tools, software and management skills are you using to navigate this crisis? We use collaboration tools like Workplace, which help employees interact with each other. We also use Asana to optimize our workflow. Our cloud-based infrastructure allows us to work from anywhere, anytime.
Who are your competitors? And how do you plan to stay in the game?
Pierre Baqué: We are unique in our field and do not have direct competitors offering similar technologies. No engineering software based on Machine Learning today would be able to process shapes directly and requires a parameterization of the design instead of its original mesh representation, nor use customers ́historical simulations as NCS does.
Your final thoughts?
Pierre Baqué: If you are flexible enough, even a pandemic cannot pose a problem to you.