syte: Helping to bring new housing to market with the speed of AI
About syte
syte is a property tech platform that's transforming real estate with AI, by making relevant property data available in real time. This allows real estate developers, investors, and architects to rapidly capture the full building potential of a property, optimize energy consumption and re-densify existing buildings to provide additional housing space.
Tell us your challenge. We're here to help.
Contact usGerman property tech startup syte uses Google Cloud tools to underpin its data-rich AI platform, accelerating the optimization of real estate development in a market with a major housing shortage.
Google Cloud results
- Maximum two seconds search response time, down from 60 seconds
- Core data processed within a week, instead of a year
- ~40% improved productivity in deploying web applications with Google Kubernetes Engine
Search response times drop from one minute to one second
Germany is facing a challenging housing shortage, with the cost of building materials skyrocketing, rising rents, and incomes squeezed by the cost of living. New studies have found that there is currently a shortage of 700,000 homes in Germany.
Re-densification—redeveloping and adding stories onto existing buildings—could be one solution to this, enabling an expansion of inner-city living without the need for further soil sealing. It's an approach that also consumes much less energy than a new build. But until recently, in order to progress building changes, architects and developers had to jump through a lot of logistical hoops, such as checking building law.
AI-driven property platform syte allows the retrieval of all relevant characteristic data on properties and their development, expansion and conversion potential in real time, making it easy to identify sites and buildings for re-densification. Using traditional methods, it would take several working days or even weeks for an architectural office to do this.
syte was born from a chance conversation between co-founders David Nellessen and Matthias Zühlke when the pair, who had met at school, ran into each other while traveling in Asia. When they spoke about Zühlke's work as an architect, Nellessen was shocked at how many processes were still being done in a very manual way. Together, they wanted to create a new way of working using AI.
By analyzing the size of an object, as well as the composition of building materials in the specific context of that object's location, syte can determine the energy requirements of a new building, as well as a re-densified area. With satellite data, it can also analyze information on the condition of the building fabric, including the materials used.
syte's AI search engine lets users search within a database of 25 million properties, filtering by dozens of filter parameters. Alongside multiple spatial queries, this search application was a big challenge for the company's database. syte needed a cloud provider with a strong focus on data processing and powerful machine learning tools.
Scaling rapidly with BigQuery
Nellessen, who is also syte's CTO, had successfully used Google Cloud in his previous startup. "Especially for startups, Google Cloud is ideal—it offers a wide range of services for every use case, but comes with a very good UX and low entry barrier compared to other large cloud providers," Nellessen says.
"Especially for startups, Google Cloud is ideal—it offers a wide range of services for every use case, but comes with a very good UX and low entry barrier compared to other large cloud providers."
—David Nellessen, Co-Founder and CTO, sytesyte has used Google Cloud since its launch in 2021, deploying BigQuery both for its data pipelines and to power its search engine. Making that search scalable and responsive was a real challenge for syte. Initially, it tried a search engine implementation with the open-source database Postgres, but found that for large spatial queries in particular, processing was slow. After switching to BigQuery, syte saw response times expedited significantly. "It was 60 seconds for the old implementation, and one or two seconds maximum with the BigQuery implementation," says Nellessen. "Our developers were really blown away by this performance boost."
All of syte's data is loaded first into a data lake, and then into BigQuery. "We're structuring the data and we're representing it all in big tables," says Nellessen. "Then there are many transformations run over these tables to enrich the data, to transform it to run the machine learning models. Our data goes through many transformation iterations, and BigQuery is really an amazing solution, because it's a fully scalable database."
"It was 60 seconds for the old implementation, and one or two seconds maximum with the BigQuery implementation. Our developers were really blown away by this performance boost."
—David Nellessen, Co-Founder and CTO, syteFinding the right piece to solve every puzzle
syte also relies on Google Kubernetes Engine (GKE), which Nellessen calls "the best Kubernetes product in the market and the backbone for a modern web application infrastructure setup." At syte, it's used for training machine learning models, running data pipelines, and hosting web services and applications. Nellessen says operating web applications with GKE is easy to set up and requires little maintenance, which is a notable efficiency booster. The team has seen up to 40% improvement in its productivity, thanks to the simplification of deploying, orchestrating, scaling, and monitoring applications.
Underpinning GKE is Compute Engine, which showed its mettle when syte was building a key data pipeline to process massive amounts of LiDAR points covering the entire country. "It's really process intensive," says Nellessen. "There are 22 terrabytes just for one federal state in Germany, and every single point needs to be classified. We first ran it on a smaller cluster, and figured out that it would take us more than a year to process, and we had to release that line in one week." By scaling up with Compute Engine, syte was able to run 600 virtual machines and process the data in parallel within a week.
Nellessen and his team value having a central console that allows simple access to a range of Google Cloud services, such as the logging and monitoring infrastructures. This allows for the easy filtering and analysis of specific logs. From the beginning, this made implementation a breeze: "When you get started. If you log in into the cloud console, it's very well documented," Nellessen says. "All the documentation is in the same style and very well explained. It's just easier to get things started."
As a software-as-a-service provider, syte can now integrate new features into the software at any time. It's able to offer those features to larger target groups with less technical effort, allowing the company to focus its efforts on key market drivers. Among those is the need to lower carbon impact in a country where the building sector is responsible for 30% of Co2 emissions. The company has already built a machine learning model to predict the energy demand for all buildings in Germany, in order to support better energy management.
"As a tech startup in the AI space, the cloud provider is key to our success," says Nellessen. "Without Google Cloud, we could not have built our product as quickly and efficiently as we did."
"As a tech startup in the AI space, the cloud provider is key to our success. Without Google Cloud, we could not have built our product as quickly and efficiently as we did."
—David Nellessen, Co-Founder and CTO, syteTell us your challenge. We're here to help.
Contact us
About syte
syte is a property tech platform that's transforming real estate with AI, by making relevant property data available in real time. This allows real estate developers, investors, and architects to rapidly capture the full building potential of a property, optimize energy consumption and re-densify existing buildings to provide additional housing space.