Learn more about

Bringing AI to the Construction Industry with EARTHBRAIN

Paul is the AI Lead at EARTHBRAIN, where he’s been tasked with building out their AI roadmap and AI team. He talks about the opportunities for AI technologies in the construction industry, what sort of team he hopes to build and what projects he expects to be working on.

Photo of Paul Neculoiu
Paul Neculoiu

Paul’s desire to advance AI development in Japan, and escape the dreary weather of northern Europe, led him to Tokyo in 2017. He said, “I first came to Japan in 2009 as an exchange student, and I moved back in late 2017. At the time, Japan was in a good position to advance AI, mainly because it was so far behind everyone else, while Europe was stagnating a little bit due to its chosen tech stack. I messaged some companies in Japan, and the one that I eventually worked for replied, so I packed my stuff and moved.”

I like seeing how information gets processed into something useful. Like, how can the machine figure it out? For me, AI gives an exciting insight into the fundamentals of thinking.

Paul has worked across various domains within AI, including neural networks and language modelling, and has also lent his expertise to various fields such as robotics, the European Space Agency and the medical industry. He said, “My background was in engineering, but then I got into robots. I realised that the maths that makes robots move is actually very complicated, but I could use a lot of simple maths and compute cycles to approximate it. Then I realised, maybe AI can fix my maths problem. So I studied AI. I did my Master’s in AI back in 2009, before AI was popular.”

In his extended tenure in the field, Paul has seen different trends come and go, and contributed to them. He said, “After I graduated, I worked for a while for the European Space Agency, working on future potential technology that might be applicable for space. You could say that I got the Agency started on neural networks. Then I got into language modelling, at a company that did CV parsing software. This was back in 2014, and all the fancy tools we have now did not exist. So, I was set about to create some of them, and we had some moderate success. Even wrote a paper on it, which got quite popular.”

Paul recently joined EARTHBRAIN, a company that provides SaaS solutions to the construction industry, helping them streamline their processes, reduce costs and increase efficiency. Currently, EARTHBRAIN successfully offers various services such as visualization tools and document and fleet management. Paul said, “The construction field is a very traditional industry, construction companies have their own established way of doing things. What EarthBrain aims to do is to help digitize and bring these industries into a more modern era.”

For example, we help construction companies fly drones over sites and digitize everything. Then based on this we can help them figure out how many construction vehicles they need, decreasing the price they have to bid.

EARTHBRAIN is now looking at how they can improve their products by incorporating AI technologies, and Paul was brought on board to help with this as their new AI Lead. He said, “On the surface it looked like there wasn’t much use for my expertise. Construction is not that popular of an industry for AI, but it actually has many pain points that could potentially be AI-ed away. My task is to try to bring EARTHBRAIN up to speed with AI technologies, figure out where these AI technologies can fit into EARTHBRAIN’s products, come up with a roadmap, and then drive the implementation of it.”

There are a bunch of hard challenges that do not have a clear answer from an AI perspective. You need to be creative in order to figure out some resolution to these problems. That’s the fun part for me.

As an AI Generalist, Paul anticipates many opportunities to delve into different fields of AI at EARTHBRAIN. He said, “There’s the whole project with machine allocation for the reinforcement learning and evolutionary algorithms. There’s customer service enhancement for all the voice, speech recognition and language model people. There’s computer vision, there’s also point cloud data, and there might be some generative AI stuff for some tasks. There’s always something new to do that just has a construction sticker on it.”

The exact projects are still under discussion, but Paul outlined some projects he expects the team will be looking into. He said, “One of the things we’ll start working on is helping construction companies with their biddings. Right now, there’s one guy that compiles the costs in a giant Excel sheet that takes him two months to do everything. If we had the right information, an algorithm might do it in five minutes. Also, soil distribution, when you have a construction site, you need to move dirt from one place to another, and how do you go about doing that in a way that is fast, cheap and safe? We want to use some AI to figure out a better solution. Other projects might be self-driving construction vehicles, like self-driving cars but bigger.”

My ideal team has people that are smarter than me and can tell me when I’m wrong.

Paul has been tasked with building out the AI team at EARTHBRAIN, and is currently recruiting for it. He said, “For now, I’m the only one. I’m mainly spending my time figuring out what EarthBrain is doing and what we should be doing with AI, and also hiring people; the interview processes, writing job descriptions, talking to people, see what kind of profile we need, who might fit the profiles. And then go about on the quest to find the right people for the future.”

The first members will have the opportunity to bring their expertise and help define the team’s processes and practices. Paul said, “I like people who have initiative and have an AI vision of their own. Software development has its own rhythm and needs. AI, I believe, has a somewhat different rhythm, the research work has a harder time being boxed up into clear sprints with clear deliverables. Generally at EARTHBRAIN each product is managed by a team, but the AI team would also be horizontally integrated with some of the other teams. We mainly provide support and create tools for other teams to integrate into the products, rather than work hand in hand with them.”

While EARTHBRAIN is a Japanese company, the team will be working in English. Paul said, “The people on my team will mainly be communicating in English because my Japanese is relative garbage, people need to be able to work with me above all. EARTHBRAIN is just at the start of its internationalization journey. I’m helping with providing feedback on what kind of foreigner support we might need to establish. Working with me, I try to advocate for the interests of people working with me. I’ve been sitting at the discussions with the management and HR for policy changes, though it can be a slow process.”

The team will be based in EARTHBRAIN’s Tokyo office, with a hybrid work style. The office itself is located in Roppongi 1-chome, about 10 seconds walk undercover from the station, which Paul appreciates after years of Europe’s rainy weather. He said, “We don’t have a ‘must be in the office for X days a week’ kind of policy, it revolves around what is the most effective for you and the team. Like if I need you to come for a discussion which is better live, then you come. Also the schedule is pretty flexible, you’re free to manage your time, as it best suits you, as long as the work gets done. Great freedom comes with responsibility.”

My policy generally is to make sure that people work on the things that they actually care about as much as possible. If you don’t care about the work that you’re doing, you’re not doing your best work.

As a manager, Paul describes himself as fairly hands-off, preferring to give his team plenty of autonomy in how they work. He believes it’s his job to understand what needs to be done and convey those requirements clearly, make sure everyone has the resources needed to do their job effectively, and then let the team do what needs doing. He said, “I like to let people have as much autonomy as they can have in order to handle the project. That being said, you’re not just left out there with no structure, guidance or support to fend for yourself. My main priority is the result, so we will prioritize whatever needs to be done in order to achieve the outcome.”