Blog / #industry

Machine Design meets Cloud Architecture and Artificial Intelligence

An interview with Max Windisch, CTO at Vention.

Etienne Lacroix CEO & Founder / Dec 30th, 2016
Vention CTO, Max Windisch

Vention is halfway between a CAD and industrial hardware company. Their platform is geared towards manufacturing engineers, enabling them to 3D design and order industrial equipment from their web browser in just a few days. For the people at Vention, the “machines that build the machines” is their niche and they intend to enter this industrial market with a business model that has an IKEA flavor combined with artificial intelligence.

As the startup is progressing through various elimination rounds of the Creative Destruction Lab in Toronto, we sat down with Max Windisch, Vention’s Chief Technology Officer, to understand how Vention can create a meaningful impact on the multi-billion dollar machine design market.

Q: What is the chain of events that led you to join Vention as CTO?

A: A short answer is that André Gauthier, a highly-respected scientist and artisan of the 3D scene in Montreal (and very good friend of mine), introduced Etienne & I to one another early summer of 2016. Etienne had been developing the idea for a few months already and had left McKinsey to pursue the idea full-time. After two months of collaboration, including several client and investor meetings, I joined full-time as CTO.

That said, many other threads weave the connection between Etienne, Vention, and myself. As a child I was a Meccano™ and Lego™ enthusiast, like Etienne. My first job in the early 90's, at Famic Inc., was to create an industrial automation software; my six years at Softimage exposed me to the demands of very high-end systems targeting teams of professionals dealing with large assemblies of 2D and 3D resources; the PI/Decho period of my life introduced me to many aspects of cloud-based and large-scale systems; and my last few years, including a brief passage at GE (in the same business unit where Etienne had worked a few years earlier), reconnected me with automation in its newer, trendier form, often referred to as the "Internet of Things".

Q: How did you use your experience to navigate the CAD field and find a niche for Vention?

A: CAD has many similarities with what I had witnessed in the 3D games & movie industry, of course. Additionally, mechanical engineering and robotics are exciting for their very neat formalism, and pretty old, robust science. There are solid ties with older branches of computer science as well (constraint satisfaction problems, planning problems, etc.), and a good amount of shared interest with more current aspects of computer vision and AI (in terms of geometric reasoning, for example).

Under extremely stringent time constraints, but with a well-defined goal to only focus on assemblies, we quickly discovered an interesting path we could navigate that seemed to have been relatively unexplored to date. There was room to develop "reasoning" at a higher level (above all that powerful geometric groundwork). For example, I noticed that an assembly containing wheels could hardly be searched for its ability to roll. When inserting parts, the most common joints were not always presented efficiently to the user. Another example is the part-whole structure of assemblies, often relying on file names and brittle file-system dependencies. From my experience, I knew that the 3D game and movie industry had developed better solutions, out of necessity, to rebuild complex wholes from multiple parts (in that industry, it's common for large teams of specialised artists to bring together enormous amounts of heterogeneous resources such as animations, characters, rigs, textures, shading etc., to produce each clip). I wanted our "3D design tool" to develop a similar mindset with respect to collaboration. With Vention, we wanted to bring classification and search to assist machine design. In the future, additional enhancements may leverage machine learning and artificial intelligence.

Q: What can you share about the features of your upcoming web-based CAD?

A: First, I would prefer not calling it a CAD. Decades of hard work of too many people have gone into CAD to pretend that we could recreate a comparable system in such a small amount of time. Our offering is very pragmatic and focused. It's based on the notion of a curated library of parts.  The "CAD" we add to this mix is only concerned with assembling parts from the library (which have previously been modelled by us behind the scenes, using full-fledged CAD like Onshape).

By offering our "3D assembler" freely, we hope to make mechanical design accessible to larger audiences, that may not have otherwise engaged in that process as much. For that reason, we feel obliged to offer a very simple, uncluttered, self-explanatory user interface.

The features we have emphasised are mostly around ease of use and speed of assembly. In most modern CADs, various smart connector offerings still require quite a bit of preparation, camera tweaking, and several feature selections. We like to present our goal as: "insert one-part in one-click". Of course, this is not always strictly possible, but with the help of AI-based ranking strategies, we're getting pretty close, and it's quite exciting.

More generally, since we control the part library, our idea is to complement it with the best online catalogue out there, crammed with useful meta-data and knowledge about the parts themselves, their characteristics, usefulness, interconnections with other parts, visualisations etc. An ordinary person, even a child, should be able to learn quite a bit about the craftsmanship of mechanical engineering, just by practising on our site. We don't want to limit the intelligence of our platform to only parts. That's why we also invest in the classification of mechanical assemblies. When you build a frame on four wheels, our system will classify it along with other similar mobile assemblies. Just like the founders of the Web envisioned a day where machines could talk to one another and interpret arbitrarily complex information (I'm thinking about RDF and the "semantic web", but also of web services, REST-ful services, collaborative systems such as Github, or the excellent Stackoverflow), we are thinking of a day not too far in the future where designers will have all the required resources at their fingertips, to reason about their mechanical designs, test them, find appropriate resources, share them or retrieve other similar designs, and enjoy that whole creative process with as little interference as possible, like children sitting in front of their set of Meccano™.

Q: What are the implications of Vention's technology on machine design?

A: We like to think that Vention will help further democratise mechanical design. Let me attempt an analogy with software. When I was a kid, it was not uncommon to work with "assembler" or "machine" language, that is, to talk to the machine directly in terms of its hardware architecture of registers, i/o, and instruction opcodes. This was not hard, and sometimes very efficient but somewhat inflexible, and it implied repeated work. In the 90's, a large proportion of serious software development was in C/C++, which was more synthetic and quite generic, yet somewhat unnecessarily cumbersome and error-prone. In the 2000s, that work started to shift toward higher-level languages. This was partly thanks to accumulated hardware and software advances that had finally rendered those solutions fast enough. It also became easier to offload work to a sizeable centralised cloud infrastructure. Today, it has come to a point where those high-level languages can take care of a huge proportion of serious software projects, leaving only very specialised portions to the older, more powerful, but more tedious approaches.

We think the same should be true of machine design. A child can produce simple mechanisms with Lego™, Meccano™, Knex™, and the like. In terms of control and automation, the theory has become widely available. It also became extremely easy to connect sensors and actuators and interact with the environment, using all sorts of cheap and convenient platforms such as Arduino. Just by making the right editorial choices, offering the right catalogue of parts, and the right tools to put together and share virtual and physical machines, we hope Vention will help bring machine design on a path comparable with the progression seen in software. This may be accomplished by revisiting the separation of roles, investing in better user experience, and combining advances from other fields such as search and AI. We hope Vention will help pave the way for more progress in that direction.

Max Windisch was interviewed on December 30, from Vention's headquarters in Montreal

Comments (0)