How We Compressed Point Clouds 20x to Stream in a Browser
There's no magic scan-to-BIM button. This episode covers reality capture, digital twins and where AI actually helps turn scans into decisions. Reality capture has quietly moved from a specialist niche to something running behind thousands of construction projects. Yet most teams still struggle to answer a deceptively simple question: does what we actually built match what we designed? Getting that answer fast, and sharing it with everyone who needs it, turns out to be where the real money is won or lost. In this episode, Dominique Pouliquen, co-founder and CEO of Cintoo, traces the path from selling his first company to build a platform that streams massive point clouds into any web browser. The conversation is a rare look inside how a category-defining product actually gets made.
0:00
Andrea Ciaffi
Hi, Dominique, welcome to the Digital Construction Podcast.
0:04
Dominique Pouliquen
Oh, hi, Andrea. Thank you for having me here today.
0:08
Andrea Ciaffi
In this episode, we sit down with Dominic Pouriquen. He's the co-founder and CEO of Sin2, the reality capture platform that turns massive point cloud data into high-fidelity 3D that any stakeholder can open in a browser. Also, Dominic led reality capture at Autodesk before founding Sin2, I'm sure many of you have seen to have used it in the past.
0:29
Andrea Ciaffi
It's a great platform. It helps really understand the difference between the 3D models and what's actually happening outside in the real world. So great tool. I'm very excited to speak to you, Dominic, today. Before we start, I have an icebreaker, something I ask everyone that joined this podcast.
0:42
Andrea Ciaffi
It's a little bit of a tax Everybody has to pay. What is the best piece of advice that you have ever received in your career?
0:56
Dominique Pouliquen
I would say the best piece of advice I kind of experience myself and I can give to any entrepreneur is multiply everything by three. I call it the pi factor. When you believe that it would take
1:09
Andrea Ciaffi
Mm.
1:14
Dominique Pouliquen
six months to get a new product, it will probably take 18 months to the market. Okay, when you believe it takes one year to get a product, it probably takes 3 years. When you believe it takes 3 months to get a new funding, probably would take nine months.
1:32
Dominique Pouliquen
You see what I mean? So always keep in mind that everything takes longer than expected, and usually it's three times longer than what you think. That means in terms of cash, in terms of visibility over your teams, think about it. all the time.
1:51
Dominique Pouliquen
Okay, there is always a big risk on this delays and usually the delay is 3 times more time than you believe it will take.
2:03
Andrea Ciaffi
Yeah, I love this piece of advice. It's very practical and I can see,
2:07
Dominique Pouliquen
You know.
2:07
Andrea Ciaffi
if you think about it, like it's as opposite as what people like Elon Musk say. He says, oh, you want to do this thing in 10 years and then think about how can you actually do it in six months. And then that's probably not reality.
2:19
Andrea Ciaffi
That's probably marketing. Yeah.
2:22
Dominique Pouliquen
Yeah. Yes, probably, yes, yes. Everything takes longer, you know, producing a new software, going to the market with a new feature, having adoption from the customers, and again, everything around funding as well. When you need to raise new funds, it always takes longer.
2:39
Dominique Pouliquen
So if you don't have the cash to sustain this period, Be careful, OK? That's that will be my piece of advice, or that's the one I have experimented many times.
2:51
Andrea Ciaffi
Yeah, definitely. And let's talk about the Sintel story. I think it's an interesting one because many people use the product, use the platform, but maybe they don't know where it's coming from. And I think it's interesting to know more about your story. You were leading reality capture at Autodesk, so you were really exposed to the problem.
3:12
Andrea Ciaffi
But at some point you decided, okay, I need to leave Autodesk and leave my career, my secure role to work on Sintu. Can you share with us what that moment felt like and why did you decide to make this leap?
3:26
Dominique Pouliquen
Yeah, no, this is absolutely. So first, I joined Autodesk through the acquisition of my first company, which was already around reality capture technologies, like photogrammetry, the technology behind, you know, you know, everything extracting 3D from 2D pictures that you use now for drone data.
3:48
Dominique Pouliquen
And my company was one of the first companies developing this kind of technology. It was also, it was also, it was a spin-off from a lab, so my co-founder was a PhD. So we run this company, it was called RealViz, and I sold it to Autodesk.
4:00
Dominique Pouliquen
Okay. So this is how I joined Autodesk. And then they asked me to join the headquarters in San Francisco. which I did to co-lead a group called Recap, the Recap team, and to build this piece of software that will manage this huge point cloud data.
4:20
Dominique Pouliquen
And I think we did a good job at Autodesk democratizing access to point clouds, you know, for all kinds of scan to BIM, you know, putting also a point cloud viewer. not only in the recap application, but also in Revit, in AutoCAD, in Civil 3D.
4:37
Dominique Pouliquen
So we put the engine in all this software so that you can easily get your point cloud data to do something with it. Okay.
4:48
Dominique Pouliquen
The big, I was lucky when I was in San Francisco to meet a new other PhDs, also from another lab here in France, that were developing a compression technology that will make this point cloud data much, you know, much more compressed so that you could eventually upload your point cloud to the cloud.
5:06
Dominique Pouliquen
Because at that time, when I was at Autodesk with managing recap, The big headache we had is that, what do I do with this huge point cloud? It was a very desktop-centric workflow. That was about 10 years ago. So uploading a huge point cloud to the cloud was very, very difficult.
5:28
Dominique Pouliquen
So there is no easy way to share the data via the cloud, to collaborate, to view the data in a web browser. And so those PhDs that came to me as Autodesk, okay, say, okay, what about this compression technology that will help, you know, compress the data so that we can easily upload the data to the cloud and make it accessible and shareable?
5:50
Dominique Pouliquen
So this is how I met my co-founders, and we decided, the three of us,
5:53
Andrea Ciaffi
And.
5:56
Dominique Pouliquen
that we'd better develop a platform ourselves than inside Autodesk, which is a big company. Autodesk was not interested by buying this technology. Okay, so I kind of quit Autodesk. When I met those guys, I first put my own money as a business angel, and then we started the company like that.
6:16
Dominique Pouliquen
They already started the company initially, but it was a technology company with no real business. Okay, so what the big decision was to make, okay, let's use this core technology that you are providing. and build a platform, okay, that we will sell to our customer.
6:33
Dominique Pouliquen
And that was really very interesting because it was solving the big headache we had with Recap at that time, which was very desktop-centric. Again, how can I make this point cloud easily accessible and shareable? via the cloud. So this is how it started.
6:49
Dominique Pouliquen
That was about 2017, roughly. Okay.
6:55
Andrea Ciaffi
Yeah, and we were talking about this in 20,
6:57
Dominique Pouliquen
Okay.
6:58
Andrea Ciaffi
the second-half of 2026, and now talking about product like platforms like Sintu sounds like obvious, right? Of course, we, of course,
7:05
Dominique Pouliquen
Yeah.
7:07
Andrea Ciaffi
we have something like that. But back in the days when you were starting, it wasn't that obvious. And I guess
7:11
Dominique Pouliquen
No.
7:15
Andrea Ciaffi
The question I have to I want to ask you is, why were you confident that the industry will eventually need a platform like this?
7:24
Dominique Pouliquen
Oh, because there was at that time, again, thanks to recap onto the product we developed at Autodesk, there was a massive democratization of the use of point cloud. And there are, you know, companies like Leica, Faro, Trimble, Topcom, developing hardware that was becoming cheaper and cheaper.
7:44
Dominique Pouliquen
I don't know if you remember, but about eight years ago, Leica release of GLK 360, that was a big stuff. So there was a democratization of laser scanning hardware as well. Okay, so it was obvious that there would be massive amount of point cloud being produced.
8:05
Dominique Pouliquen
But if you needed to send a hard disk. to your teams or to your customers to share this data, that was not, you could not have a big adoption of this point cloud, okay? So at that time as well, you know, the cloud, so everything in the cloud was booming, okay?
8:20
Dominique Pouliquen
And so yeah, the connection between the two was obvious, okay? If you make your... huge point cloud, usually, we're using the term cloud for two issues here, two topics. First is a point cloud, the one, the data you get from the laser scanner, and we use a cloud,
8:42
Andrea Ciaffi
Yeah.
8:44
Dominique Pouliquen
which is a hosting platform, you know, in the on data centers.
8:45
Andrea Ciaffi
Yeah.
8:50
Dominique Pouliquen
So yeah, putting a point cloud in the cloud would make this data fully, easily accessible, manageable, and being able to stream this data in a simple web browser with no plugin to install, so easy, you know, easy setup, was, I was obviously, it would It would be a big adoption, which which it was OK, so yeah, this is why we I felt that all the all the customers that already had recap that already had bought a laser scanner would be interested by this platform.
9:26
Andrea Ciaffi
Yeah, it's funny, the word cloud will be the keyword of the
9:29
Dominique Pouliquen
Yeah, exactly. It's cloudy.
9:30
Andrea Ciaffi
day. Yeah, yeah. And every time somebody tries to start a
9:32
Dominique Pouliquen
Exactly.
9:34
Andrea Ciaffi
new platform, a new company, there are things you, at the beginning, you get right, and that's exciting, and you build your business, but sometimes there are things that you learn from from your customers, where your plan, it's the real world, and now you're confronted with the reality.
9:49
Andrea Ciaffi
Is there anything, what are the things that from the beginning you found, okay, this was my assumption, my thesis, and it worked? And what are the things that kind of surprised you when you see the customers were taking on different angle to the problem?
10:07
Dominique Pouliquen
That's a very interesting. So the core technology we got from my co-founders, PhDs, was this capacity to turn a huge point cloud into a compressed data. The way they do that is by turning that into 3D mesh or 3D surfaces. And that will solve another problem we had with point cloud.
10:25
Dominique Pouliquen
Is the point cloud is like a 3D dots in space. So you see through the dots, okay? It's not continuous information. But the world we are looking at is made of surfaces, okay? You don't see through the walls, okay? So by turning point cloud into 3D surfaces, I was also pretty convinced that this would be a... much easier way to interpret your point cloud data.
10:49
Dominique Pouliquen
Okay, so 2 aspects here. First, we could compress it by turning it into meshes. Secondly, the 3D surface you get or the 3D mesh you get will make the data much easier to view and interpret.
11:05
Dominique Pouliquen
And last but not least, the core technology also includes a fast, the fact that we can stream these high resolution 3D meshes, which are at the resolution of the source scan, in a web browser using fast streaming. So people at that time were calling us, oh, you have the Netflix for 3D data, because we could stream high resolution 3D meshes in a web browser.
11:29
Andrea Ciaffi
That's funny.
11:30
Dominique Pouliquen
Okay, that was really cool. And the technology is so advanced that once we turn a point cloud into a mesh, we can do the inverse transformation, mesh back to the source point cloud with no loss of data. Okay, almost no loss. Okay.
11:45
Dominique Pouliquen
And that's really interesting because in some use cases, When you need to have your reality capture data in Revit for scan to BIM, you need a point cloud. Because what we did with the recap team at that time, with the Revit team, by putting the recap engine into Revit, you would be able to snap primitives of Revit to the point cloud.
12:08
Dominique Pouliquen
So you need point cloud in Revit, okay? Now, it's kind of changing. Now, people can use meshes also in Revit, so, but that was that was the initial value proposition we had with Scene 2. OK, so compression, we could compress the data 20 times.
12:22
Dominique Pouliquen
OK, so from 20 to one, we could upload then the data to the cloud, we could stream very fast. and we could view these beautiful meshes in a web browser. Or we could turn back the mesh into point cloud when you needed the data back in Revit.
12:42
Andrea Ciaffi
Yeah, definitely a brilliant solution. But if you think about any platform,
12:44
Dominique Pouliquen
Yeah.
12:47
Andrea Ciaffi
any company, at the end of the day, you need customers and you have great customers. Like you have very large companies like
12:53
Dominique Pouliquen
Yeah.
12:55
Andrea Ciaffi
Arup. I worked at Arup for seven years. Stantec, Talent Construction,
12:55
Dominique Pouliquen
Mhm.
12:59
Andrea Ciaffi
General Motors, just to name a few. But I think it's interesting to understand where do these customers actually see the value of the platform? What is the value actually coming from? Is it the savings on projects? What is it that seem to bring?
13:16
Andrea Ciaffi
And do you think that the value you are expecting to provide to your customer is the same value that your customers actually experienced from the platform today.
13:27
Dominique Pouliquen
That's an interesting question, but before going there, when we started also Sintu as a platform, the key question was a business model. Okay. How can we make this platform easily adopted by those big companies? Okay. And we kind of navigate, you know, it's is the role of a startup to select what we decided from the beginning.
13:53
Dominique Pouliquen
That was a key decision. We started that we would be, our policy would be for unlimited users. So we would not base our business model on the number of users. But on the quantity of data that you would upload. And I think this is an interesting, you know, that was a good move.
14:16
Dominique Pouliquen
That was a good decision we made because when you don't have to question about, okay, how many users, whatever, we say, we always say to our customers, please invite as many users as you can. Okay. Don't worry, it could be your users, you could be your contractors, you could be your customers.
14:36
Dominique Pouliquen
Feel free to invite as many as you want. On today, we have more than 110, 000 active users on the platform. So that was number one. The #2 business model decision we made at the beginning or scale, we would base on conception, based on tokens.
14:55
Dominique Pouliquen
Okay, as Autodesk was starting to do. And that was not a good decision. So we had to pivot a little bit because we found that in those companies you mentioned, Arab, Turner Construction, whoever, the buyer and the user are not the same, of course.
15:09
Dominique Pouliquen
Okay. And when you buy tokens, the users sometimes feel afraid of consuming the tokens because they believe they consume, you know, cash of the company, you see. And they are kind of keen, oh, should I consume those tokens and then I will be, my manager will not be happy because I consume too many tokens, you see.
15:33
Dominique Pouliquen
So Based on managing tokens was not a good decision, so we pivot. just purely on the size of the data. So you say you have a project for, I don't know, a big project with 10, 000 scans on a big facility, please subscribe to 10,000 scans, okay?
15:51
Dominique Pouliquen
Okay, but so not based on consumption, based on the data quantity uploaded to the platform. And we still, we still use this business model. OK, and that was that was a good decision.
16:09
Andrea Ciaffi
Yeah, that's much easier to understand. Yeah. Gary hard.
16:09
Dominique Pouliquen
So. Yeah, much easier, much easier to, and then people will start, eventually will start small, they can buy a subscription for, I don't know, 2000 scans, 5000 scans, and then when they need more scan capacity, they just upgrade easily. Okay, so that was our business model, and this still is our business model.
16:33
Dominique Pouliquen
Revenue today is coming from, you know, first subscriptions. But because the customers love the platform,
16:38
Andrea Ciaffi
Yeah.
16:42
Dominique Pouliquen
we have very high retention rate. We also have a very good success in upselling, so in upgrading the capacity of our customers.
16:54
Andrea Ciaffi
Yeah, and...
16:54
Dominique Pouliquen
So, these all...
16:56
Andrea Ciaffi
One of the things I clearly remember the first time I saw Sintu in a project. And one of the things that impressed me about Sintu was how easy it was to show 3D data to people that didn't know anything about digital. And I was like, oh, wow, I'm looking at this train station.
17:15
Andrea Ciaffi
With all of the different, so it was a it was a Revit model of a train station in Melbourne that was going to be built, and we put the Revit model, the the data from the data from the DB scan everything together, and that felt so easy to show the data to people that are not necessarily specialists.
17:34
Andrea Ciaffi
of this of this field. Now, that sounds simple to put everything
17:36
Dominique Pouliquen
Yeah, yeah.
17:39
Andrea Ciaffi
together when we discuss today, but I think most of the product they
17:41
Dominique Pouliquen
Yeah.
17:43
Andrea Ciaffi
still fail at bringing everything together and showing the data in a way that is easy to understand for non-specialists. Why do you think this step is so hard to do and it's so much harder than probably
17:52
Dominique Pouliquen
Yeah.
17:57
Andrea Ciaffi
looks?
17:59
Dominique Pouliquen
Ohh, there there are many good reasons, so what what we do at at Sin Two, I think we do a pretty good job at that is... putting our own feet in the customer shoes, which means, okay, what as a user, what would I like?
18:13
Dominique Pouliquen
So the user experience, we have people that are specialists in user experience in Sim2 that tell us, okay, you should do it this way and this way. So I think that still in Sim2 from the beginning, the user experience has been really key.
18:28
Dominique Pouliquen
The way you view the data, the way you organize the data, you organize in work zones and everything. And of course, what we said from the very beginning is that it's not only for point cloud. What you want to bring inside SIM2 is your as-built reality, which is a reality capture data, and you want to bring your BIM model and you want to compare the two.
18:50
Dominique Pouliquen
So from the beginning. one of the key workflow was QAQC. How can I compare my existing conditions to the design intent? So from the beginning, we connected SIM2 to Autodesk Construction Cloud. At that time, it was called BIM 360 or this kind of stuff.
19:05
Dominique Pouliquen
So cloud to cloud, you can bring in Revit models, Navisworks models, AutoCAD models. 2D and 3D, okay, with all the metadata that comes with it. So you have all the disciplines on everything. So you have inside scene 2 a complete replica of your CAD model, okay?
19:26
Dominique Pouliquen
And then it's all about making sure they are aligned. That is a key point, okay? So we kind of solve that in a... pretty okay way we are working on a better alignment tool, but still, in most cases, people will use survey points anyway to align the point cloud to the Revit model, okay?
19:46
Dominique Pouliquen
And so this problem was solved pretty easily. And then it's about, okay, how can I view the differences between the two? And we had this very early stage red on green, you know, user experience where you will see everything green is a good match of the BIM model on the scan data within a certain tolerance.
20:08
Dominique Pouliquen
Everything red is data that is in the scan, but not in the BIM.
20:13
Dominique Pouliquen
So people would easily interpret Differences between the beam on the on the on the on the scan data, but other than that, yeah, so just the easiness of use or the user experience comes fast from the fact that you can easily stream very high resolution 3D mesh at any time, so you would navigate scan to scan very easily.
20:36
Dominique Pouliquen
or you will navigate in 3D bird's eye view, you see. What we felt in many, many occasions is that people rather navigate their facility or their scan data going from scan to scan, because it's a kind of Google Street View experience, you see.
20:52
Dominique Pouliquen
And not many people are used to flying
20:55
Andrea Ciaffi
Yeah.
20:57
Dominique Pouliquen
around because you, oh wow, you get dizzy, where am I going? So from the beginning, we said that we should have two modes, scan to scan, okay, going from a scan position to another scan position, and which we enrich later, which with a feature that we call teleport.
21:16
Dominique Pouliquen
So virtually, if you want to move from this scan position to another position that has no camera location or scan location, you virtually create a scan location here. Okay? And then, yeah, so this is a long, sorry, I'm taking too much time here, but the user experience has been always a key point for us, making sure that people can easily navigate your scan data.
21:44
Dominique Pouliquen
And the streaming aspect, the technology behind makes it really cool and very easy. Yeah.
21:51
Andrea Ciaffi
Yeah, and I agree, as a user, I've used it most of the time using scan, like scan to scan, jumping from one red dot to the other one.
21:56
Dominique Pouliquen
Yeah. Exactly, yeah.
22:00
Andrea Ciaffi
If it's a lot easier because you don't get dizzy moving the 3D world,
22:01
Dominique Pouliquen
Yeah, exactly.
22:05
Andrea Ciaffi
that's a real, it's a real improvement to the user experience.
22:09
Dominique Pouliquen
Yeah.
22:10
Andrea Ciaffi
And when you, one of the good things that you can do with scene 2 is being able to catch discrepancies before they become actual real headaches and money lost and probably people have to rebuild stuff.
22:21
Dominique Pouliquen
Yeah. Yeah.
22:24
Andrea Ciaffi
Because of course you can you can mix, as you said, like the. 3D models, which is just a project with the real world, with the point cloud. Have you seen any expensive discrepancy that customers have been able to avoid just because they've been able to find what was happening in the real life before continuing and make it big damage?
22:48
Dominique Pouliquen
Oh yeah, this is our everyday bread and butter for our customers, you see. So as you said, customers like Arab or Alterna Construction or whoever, I just thought they would expect the leadership would offer better documentation, scan to BIM efficiency, What we really provide, we do provide all of this, but what we do really provide is a shared source of truth across fragmented teams.
23:16
Dominique Pouliquen
And that's already a big, big value proposition for all these companies. And then it will allow faster decision making and just better documentation. So it's all about the speed at which you can make your decision, okay?
23:31
Dominique Pouliquen
And as I said, with this kind of UX, you know, red on green stuff that we introduced seven years, eight years ago, you have earlier visibility into a divergence between your as-built and your scans. Okay, I'm sorry, and as-built versus. to design intent.
23:50
Dominique Pouliquen
And that was really something that would, for which we gain trust from the customers. So they would easily hand over the data to their end users. They are sure that the data doesn't stay within the head of one. Virtual BIM Manager or VDC Manager is becoming industrial knowledge.
24:10
Dominique Pouliquen
OK, institute, so... We started from a, okay, let's give a better documentation tool to a platform that gives really a low, higher confidence decision making, okay, across fragmented teams. The fact that we have this unlimited policy, you know, we can easily invite as many users as you want.
24:39
Dominique Pouliquen
You can give permissions to these users. You can allow access to certain areas only. OK, so that will. We had a very massive expansion of the user base because of that. Okay. So yeah, Arab may see it as a design and coordination enhancer, but we strictly focus on better decision making against real world conditions.
25:05
Dominique Pouliquen
And Turner may see it as a project execution tool, but the real value is really about controlling risk, by making field reality visible and actionable sooner. OK, if it makes sense what I'm saying. OK.
25:21
Andrea Ciaffi
Yeah. Definitely, you can make a decision and not quickly,
25:25
Dominique Pouliquen
Yes.
25:26
Andrea Ciaffi
and it sounds it sounds simple to describe, because now seem to exist, but the ability to verify what's happening on, so there are many people,
25:31
Dominique Pouliquen
Yeah.
25:36
Andrea Ciaffi
many people working construction, they have been able to verify what's. was happening in construction even before Sinto existed,
25:44
Dominique Pouliquen
Yeah, yeah.
25:45
Andrea Ciaffi
but of course it wasn't that efficient. What do you think is the risk of making the decision? Because somebody can always make a decision and say, I'm not going to use a platform like Sinto, I'm just doing it the old way, whichever old way they pick.
26:00
Andrea Ciaffi
What type of risk do you think they can run into? by doing that the old way.
26:06
Dominique Pouliquen
Oh, it depends on use cases, okay? There are many use cases, even in construction, okay? I would tend to summarize that in two big categories. First is scan to BIM. When, you know, 60 or 70% of the business, of the construction business in the US, is about renovation.
26:25
Dominique Pouliquen
Okay, so you have to know your existing conditions. Okay, that you want to turn into a BIM model, of course, because you want to design a new design inside Revit or Navisworks or whatever, okay? So you have to turn this existing condition into a BIM model, and that's already.
26:45
Dominique Pouliquen
the core aspect of, and the dream, the dream scan to BIM button. Okay, I have my scan data, I have my point cloud, I press a button, I have my BIM model level 400. No, it doesn't exist yet. So in most cases, what we allow with SIM to is an easier workflow for modeling teams.
27:04
Dominique Pouliquen
Usually they are offshore, like in India, Vietnam. They will build the model from the point cloud, and then you can QAQC the quality of the reconstructed BIM model. And when you have a BIM model that fits your as-built conditions, then you can start a good design.
27:21
Dominique Pouliquen
So as I would say, this is my first category of use cases. The second category is about when you build, when you start building, then you want to make sure that what is being built is according to the design, to the BIM model, okay?
27:35
Dominique Pouliquen
And this is where you have this QAQC process. And it's not only when you question about which... which example you have of a big something that happened, is usually not that big. It's just a few inches off, okay? But a few inches off of people putting
27:55
Andrea Ciaffi
Yeah.
27:57
Dominique Pouliquen
the duct and people putting the pipes and putting, will, you know, at the end of... The following week will be a big headache for the next, you know, provider or the next guys coming in, so you have to find these small inches, you know, issues.
28:17
Dominique Pouliquen
along the way, just to avoid the big, big problem. And sometimes a big problem is to is to delete everything on the build again. So, you know, it's not a big, big, usually a big drama. It's just a few inches off of a few of a few, you know, people building.
28:36
Dominique Pouliquen
and that would make the big issue at the end. So yeah, this is how we started with companies like Turner Construction, IRAP. To be honest, there's been on this second category, there is a shift, to be honest. It's laser scanning cost quite a lot, although there is now mobile, you know, mobile scanning.
28:56
Dominique Pouliquen
that is coming. You can use a backpack, you can use a handheld device. They don't have the accuracy needed for QAQC though, so this is an issue. But also people use 360 cameras now, okay, to have a quick overview of what is being built versus what should be built.
29:12
Dominique Pouliquen
But yeah. So there has been a lot of new technologies being introduced for this QAQC. And what we have at SIM 2 also is a hardware agnostic and multi-modal platform. Inside SIM 2, you can bring not only point cloud data, drone data, but also now 360 videos that you capture with your camera on your helmet, okay?
29:36
Dominique Pouliquen
So all this kind of data will participate to your QAQC process.
29:41
Andrea Ciaffi
Yeah, I've seen it done in a project for a data center where it was a very complex problem because we don't have, like you said, we don't have this magic button 3D scan to BIM. And the job there was try to remodel an
29:54
Dominique Pouliquen
Yeah.
29:56
Andrea Ciaffi
entire data center. And those centers are very complex in terms of like mechanical systems.
30:01
Dominique Pouliquen
Yeah.
30:03
Andrea Ciaffi
So, and the ability to load different types of data really helped, because sometimes with point cloud, you don't understand the color of things, but being able to see the data in
30:11
Dominique Pouliquen
Yeah.
30:13
Andrea Ciaffi
different formats really helped the team deliver the project.
30:17
Dominique Pouliquen
Yeah.
30:17
Andrea Ciaffi
You have been in the reality capture for a long time, and I think this is my opportunity to ask you a couple of questions about digital
30:20
Dominique Pouliquen
Yeah.
30:25
Andrea Ciaffi
twins.
30:26
Dominique Pouliquen
Yep.
30:26
Andrea Ciaffi
Because, you know, there was a lot of hype in the past. Some people have lost a little bit of faith, faith in the digital twins. What's your take on digital twins in terms of like, where do you see digital twins delivering value versus sometimes they're just theater and...
30:46
Andrea Ciaffi
Marketing.
30:48
Dominique Pouliquen
Oh no, no, no, it's much more than that now. A digital twin. So that's a good question. But anyway, all your questions are good. This is what I know. But anyway, what we call a digital twin. I would say, or what was called a digital twin initially, was a connected CAD model or connected BIM model.
31:09
Dominique Pouliquen
So you start with a BIM model. BIM model is a digital representation of your building or your asset or whatever. It is a parametric model, so it has a lot of information contained in each In each piece of your bottle. But it's not connected, it's not a live.
31:27
Dominique Pouliquen
So if you are not connected to IoT, you are not connected to documentation platform and things like that. So when you start connecting your CAT model or your BIM model to live sensor information, to ERP system, something like that, then you are starting to create a digital twin, definitely.
31:47
Dominique Pouliquen
Okay. I would say if you add one step further and that was a promise of meta at some point, you start putting man in the loop, then you start building an industrial metaverse. You have a connected cat model with human in the loop where you can do real simulations, then you create an industrial metaverse.
32:06
Dominique Pouliquen
So as you see there, BIMs, BIM or CAD models, digital twins, and even eventually the ultimate proposition will be industrial metaverse.
32:16
Dominique Pouliquen
But coming back to the BIM, to the digital twins, there is a, so I think there is a big, big I would say shift, I wouldn't say a shift, but when we started discussing digital twins, mostly this term came from the energy sector, like oil and gas, where they have, you know, refineries, they have, you know, a big, big production facility full of equipment.
32:48
Dominique Pouliquen
You know, one oil rig is a North Sea for one of our customers, BP, has 150, 000 equipment inside. Okay. And all this equipment has need to have maintenance logs, need to be, you know, maintained, need to be documented, nothing like that. So you need, you want to create a digital twin to have a visibility of these big assets.
33:12
Dominique Pouliquen
So the initial, okay, value proposition, say, okay, let's turn back, you know, this huge facility into a BIM or a CAD model, and then we will connect it to connect it to SAP to connect it to IoT platforms, and that will be my digital twin.
33:27
Dominique Pouliquen
But it's so complex. It's so complex to remodel. these facilities. Okay, facilities that have not been, you know, you don't have an updated CAD model for many years. So when you want to remodel, so more and more people believe that the digital twin is a scan data.
33:49
Dominique Pouliquen
As long as the scan data is segmented in a way that you can isolate each piece of equipment. Okay, so why remodel? when the data is in your scans. You see what I mean? And so, yeah,
34:01
Andrea Ciaffi
Yeah.
34:02
Dominique Pouliquen
so that means that our value proposition as well has shifted a little bit from, okay, we host the data, you can compare to your scan or your scan to your BIM to, okay, let's extract meaningful information from the scans to the point that.
34:17
Dominique Pouliquen
You have enough information to consider that the scan, your segmented scan, is a digital twin that you can connect. You can connect each equipment. We create what we call tags. Okay, a tag is a bounding box that contains an equipment or a pipe, and you can connect that to either a CAD model, but also an ARP system or an IoT platform, etc. You see?
34:40
Dominique Pouliquen
So, yeah, no, it's a big move.
34:42
Andrea Ciaffi
Yeah. Yeah, that's a very interesting vision and interpretation of digital twin, because at the end of the day, things like Revit or BIM started with the idea of objects that are oriented in the space. It doesn't have to be a Revit object, could be anything.
35:03
Andrea Ciaffi
So, it's a very interesting approach you're taking and I think it's a very good one, so you don't have to remodel everything.
35:11
Dominique Pouliquen
Yeah.
35:11
Andrea Ciaffi
Very good point. In this space though, now we are also dealing with AI because of course AI is touching everything, of course, and it's probably touching even reality capture and the ability to identify things in the space. You already mentioned, we already know that we don't have this magic button, convert this scan into Revit model.
35:41
Andrea Ciaffi
What do you see AI can do today? that was not able to do, we were not able to do a couple of years ago.
35:49
Dominique Pouliquen
Well, absolutely. So there are many aspects to it. OK, the first one, the first AI we introduced in Sintu was six, seven years ago. We call it machine learning. OK, basically is OK, I have my scan data. I can see either panoramic images or my
36:00
Andrea Ciaffi
Mm.
36:05
Dominique Pouliquen
3D surface. How can I use this meaningful information? to detect equipment. So we train our machine learning engine to detect, now we can detect 50 plus classes of equipment, like valves, pumps. Of course, we also do geometric segmentation, so we can detect pipes, walls, and everything like that.
36:22
Dominique Pouliquen
Okay, so AI has been, through machine learning, a key component of value of the value proposition for many years, and we enrich this library of, you know, all the equipment we can detect. Almost every month, we have a new class being added to the library.
36:44
Dominique Pouliquen
OK, and that's one way of using AI is machine learning. The other way is agentification. You know, Sim Two comes with a complete set of APIs, SDK. So, how can I, you know, how can I interrogate my own data using Sim Two API, and how can I use AI agents inside Sim Two to automate some workflows as well?
37:08
Dominique Pouliquen
OK. like I want to export data, you can do that using scripts, you can do that, you can embed even sync to viewer into your own AI agents. So there are many ways where AI is being used by our customers today. And this will explode, okay, because you can automate a lot of workflow, you can.
37:32
Dominique Pouliquen
look for data. You have massive amount of data available through SIM tool, so you can extract meaningful information, because data is all about, you know, extracting, you know, patterns, extracting, making your life easier when you have repetitive tasks, etc. There is no magic button, please, AI, understand my job site.
37:52
Dominique Pouliquen
No, no, there is not yet, okay? Because you are missing the context,
37:58
Andrea Ciaffi
Yeah.
38:02
Dominique Pouliquen
you are missing, you know, contractual scope, you are missing, so I don't know, any kind of installation. intent and whatever, okay? So there's still a tool to help people at the end of the day, okay? But AI will allow you to classify and geolocate your equipment, eventually compare.
38:20
Dominique Pouliquen
But coming back to my, so what we do inside Sintu using AI, we create those tags which are bounding boxes, okay? And we know, okay, this is a valve. We know it's a valve of a certain type, but we don't know yet the serial number of this tag.
38:37
Dominique Pouliquen
Okay, so it's very still a manual process to connect this tag to, let's say, is maintenance log in SAP, for example, because you need to have the exact serial number of this equipment. And sometimes you don't see this information from the scan data because you don't have enough resolution.
38:59
Dominique Pouliquen
Reality capture is also a line of sight technology, so there are a lot of occlusions. You don't see all the information. So you know there is a valve, you know it's a certain type of a valve or a pump. But connecting this tag to the right serial number that will give you access to the maintenance log of this tag is still a pretty manual.
39:21
Dominique Pouliquen
You see what I mean?
39:24
Andrea Ciaffi
So, you still see the human in the loop in this in this type of jobs?
39:26
Dominique Pouliquen
Oh yeah, oh yeah, yeah, yeah. And we see that from our customers in oil
39:29
Andrea Ciaffi
Yeah.
39:32
Dominique Pouliquen
and gas or even manufacturing when they want to use into for asset management on being the source of truth to connect to their ERP system. A lot of manual work is being outsourced. Again, to even when you need to connect, for example, your 3D tags to 2D P&IDs.
39:52
Dominique Pouliquen
OK, P&IDs is still used a lot, you know, in this process, you know, process industries. OK, you can have P&IDs for your water, for your oil, for your whatever. Anyway, so connecting these 2D symbols to the 3D tags is not that obvious, and we're working on automation, but still with human in the loop, absolutely.
40:17
Andrea Ciaffi
Definitely. And I think this is another opportunity for me to ask you a question about the future. Like,
40:22
Dominique Pouliquen
Mhm.
40:23
Andrea Ciaffi
it's very hard for anyone to predict the future, but I think usually CEOs, they tend to think about the future a lot because you try to create a vision for
40:31
Dominique Pouliquen
He.
40:33
Andrea Ciaffi
your company and try to anticipate future trends. And your company lives at the intersection between reality capture, BIM, GIS, and now there is AI, throw AI in the mix. So there's so many moving parts. But if you look into the future of three to five years, what do you think becomes normal in reality capture that Most of the industry will still find surprising today.
41:01
Dominique Pouliquen
That's a really good question. There's still a lot of... problems to solve, okay? So like, for example, what before I get to this point, okay, what we see now is a massive trend to have bigger and bigger projects. For example, it's not unusual inside scene 2 that you are being asked to see a complete.
41:27
Dominique Pouliquen
compass or complete refinery. So we are still working on optimizing the viewer so that you can view, I don't know, 10,000, 15,000 scans, plus a few gigabytes of BIM model, plus 2D drawings, plus GIS layers, all of this. must be seen in a few seconds in a web browser.
41:46
Dominique Pouliquen
Can you imagine that? That was completely out of reach three, five years ago, okay?
41:53
Andrea Ciaffi
Yeah.
41:55
Dominique Pouliquen
People would not even expect that from a simple web viewer, Google Chrome or whatever. You can see this massive amount of data in less than a few seconds. And if they wait more than a few seconds, they say, oh, come on, how come it doesn't work?
42:09
Dominique Pouliquen
But what sometimes when you upload the same amount of data in a desktop software, like even Recap, you have to wait for one hour before seeing anything, you see?
42:19
Andrea Ciaffi
Yeah.
42:20
Dominique Pouliquen
And so we are being challenged all the time by our big customers to say, okay, bring in massive amount of data and we're solving that. Okay, we're solving that. This is 1 aspect. Another aspect is everything hosting. You know, we have cybersecurity issues.
42:34
Dominique Pouliquen
Of course, we are SOC 2 type 2, we're ISO 27,001. We can already provide a lot of choices for the user on data centers in any region, everything like that. But now what we see, particularly with the current administration we have in the US, people in Europe want to use providers that are in Europe.
43:01
Dominique Pouliquen
They don't trust Amazon or Azure anymore because there is a cloud act by which at any time the US government may request access to the data. Even if the data center is in Europe, it's still a US provider. So there is a trend also for data.
43:14
Dominique Pouliquen
We call it data sovereignty.
43:16
Andrea Ciaffi
Mm.
43:19
Dominique Pouliquen
So we still need to work on that. There is another aspect, okay, which is, oh, I want to be in the cloud, but completely disconnected from public clouds. So we call it self-hosting, so also working on this kind of stuff. So this is taking a lot of energy.
43:32
Dominique Pouliquen
Now what is happening for sure is that there is a massive democratization of reality capture. devices. There has been a big push of the mobile mapping. Companies like Navis, Emesent, you know, now we have people from even from China with X-rids, they do provide cheaper and cheaper and easier to use equipment.
43:57
Dominique Pouliquen
Now there has been also a big trend over the last 18 months. is Gaussian splatting. I don't know if you heard about this new way of looking at the 3D data, mostly working with photogrammetry data though, but it's also coming in the world of point cloud and laser scanning for mobile, at least for mobile mapping.
44:17
Dominique Pouliquen
So there are big trends here happening and this is good for the community. So definitely I would see that in the future, reality capture data will become continuous. Okay, you will capture all the time. There will be devices on site or there will be people going every day capturing data, more and more data, and to capture facilities so not episodic.
44:42
Dominique Pouliquen
capture, but continuous capture. And then you can, you need to compare the scan data from day one to scan data from day three, etc. That's for sure. Scan data will become also operational, not archival, okay? So you have, on AI will help certainly To turn unstructured point cloud into something that easy to easier to search on on on segment also as well, so I'm taking, yeah, so.
45:15
Dominique Pouliquen
There are probably a big trend in the fact that there are more and more reality capture data, more continuous data. Yeah, that's a way I can see that.
45:25
Andrea Ciaffi
Yeah, that's great. And we are at the end of the episode, Dominic. And I want to say I really enjoyed this conversation with you because you don't see very often a CEO that is in the weeds like you are, that knows the technical like you do,
45:37
Dominique Pouliquen
Yeah.
45:38
Andrea Ciaffi
which is very exciting. So I'm very glad that I had this
45:40
Dominique Pouliquen
Mhm.
45:41
Andrea Ciaffi
conversation with you. If people want to reach out to you, where can they find you?
45:47
Dominique Pouliquen
Best way is to use a website, sim2.com. We have many ways people can request a demo, can request a free trial. I'm happy to respond as well directly to people. They can reach me on LinkedIn. So Dominic Poliquin on LinkedIn. Pretty easy to find, or through my e-mail at pouliquen@cinto.com very simple.
46:10
Dominique Pouliquen
I'm very responsive, and if they want to have any, yeah, the best way is to use a website. There is also, of course, sales@cinto. com on contact@cinto.com, but anyway. LinkedIn for me would be the best, certainly, for me personally.
46:30
Andrea Ciaffi
Great. I will put the links into the show notes. Thank you for joining us on the Digital Construction Podcast.
46:33
Dominique Pouliquen
Yeah. Thank you. Thank you, Andrea. Thank you for hosting me. Thank you.

15 April 2025
Online Training and Digital Adoption with Susan Brattberg
Understand why a comprehensive training plan is a crucial insurance policy for technology investments