Inside Builder Knowledge Graph (BKG): The brain that powers our ecosystem

Discover how we automate the assembly line and power entities like Features and Templates.

Andreas Antoniades

Graph Machine Learning Lead at
· 4 minute read
Inside the Builder Knowledge Graph

Who is

We are on a mission to unlock everyone’s potential with the power of software! Our combined approach of AI, automation and talented humans means that your background, tech knowledge or budget will never hold you back.

Learn more about us

The world doesn’t come with a manual. When you step outside the digital world there’s no set model of how things work or where to find them. Mostly our brains describe tangible things by relating them to something else.So why should data have a set model of rows and columns that doesn’t reflect the world it’s describing?

That’s why the most successful businesses of the last decade have moved away from those models and now use graphs to store and interact with their data. From social networking platforms, to search engines, media applications and even software building platforms.

Need more app insight? Join the community

Stay up-to-date with the latest industry trends

By proceeding you agree to's privacy policy and terms and conditions

What’s a graph database?

Graphs are a way of thinking about and storing data, they’re far better at showing interconnectivity than a traditional database. We think of the Builder Knowledge Graph as the brain of our operations, because it powers most of our workflows.

Nodes (the coloured circles below) represent the entities that exist within the ecosystem – from Features, to Templates, Problems and Projects. The edges (that’s the lines that connect nodes) on a graph represent how these nodes are related. For example, a feature can be included in your project as part of a template, while a template is responsible for solving a specific problem.

Example of how the Builder Knowledge Graph is structured

Next, you need graph embeddings

To harness the power of graphs, you need to help machines to be able to read and learn from your data.

We use graph embedding algorithms to do this. They transform the structural and attribute information from a graph, into numerical n-dimensional vectors. This means that now, machine learning algorithms can read the graph.

Then the fun begins. A graph-enabled machine learning algorithm automatically learns these numerical representations – and embeds information about each one and its most relevant connections.

How we generate graph embeddings using random walks and FastRP (an algorithm that learns word associations)

So what can graphs do?

You give us a problem statement – by telling us what problem you want your app to solve. We combine natural language processing and graph-generated embeddings to recommend some project templates that could power your solution.

Architecture of the template recommendation system

Can graphs recommend the best features too?

You can create a blueprint of your app idea by adding features (that’s things like Login, Push Notifications or Barcode Scanner) to your Buildcard™. It’s like adding toppings on a pizza. The Builder Knowledge Graph uses the template embeddings and feature embeddings that it’s learnt, to give you relevant recommendations – of extra features that could improve your app.

Image: A feature is recommended using the Builder Knowledge Graph (BKG)


Going beyond the graph

But can this technology create something new? Builder Now is our app prototyping tool (it helps you design your user journeys and improve your app, by offering instant app prototypes). Again, it’s only possible thanks to the numerical representations of features and templates that our graph has learned.

The Builder Knowledge Graph uses what it’s learned to spot relevant feature-to-feature connections – allowing it to automatically create a prototype for a completely new app.

How Builder Knowledge Graph generates prototypes

Allocate work more efficiently

Historical data from all the projects we’ve ever done is fed into The Builder Knowledge Graph, which captures the complex interactions between features. These insights mean we can split a project up into parallel workstreams and have multiple people working on them at the same time – without holding each other up.

Parallel workstreams

The right person for each job, instantly

And you can use these graph embeddings for any connections that are relevant to your business. Who works on every task of all our projects is also powered by our Builder Knowledge Graph.

For every designer or developer we work with – our embeddings work out their expertise level, past experience on similar features, and which timezone they’re in. Allocating work becomes automated based on each person’s merits, removing any bias.


Selecting experts for different tasks, based on skills and experience

Spotting problems before they happen

Information from past projects stored in the Builder Knowledge Graph can work out the probability of your project being delayed too. It can also predict the most likely causes for a delay, for example, a customer is slow giving feedback, developers are raising more bugs, or feature to feature interactions are more complex. Then the Builder Knowledge Graph uses this insight to suggest actions that our project managers should take to avoid this delay.

Example Bayesian network: captures cause-effect relationships to calculate the chance of a project delay

How will you use this power?

You’ve read how our own Builder Knowledge Graph is really the brain behind operations – allowing us to do everything from suggest relevant features to you, generate instant prototypes, work out who should do each task (and in what order). Hopefully this has given you a flavour of the power of graph databases. You’ve seen how using graph embeddings makes your data readable by machine learning algorithms. After that, your possibilities are wide open.

Join the community

Stay up-to-date with the latest industry trends

By proceeding you agree to's privacy policy and terms and conditions

Andreas Antoniades

Graph Machine Learning Lead at

Dr. Andreas Antoniades is the graph machine learning lead at where he improves technology using his knowledge of data science, graphs and neural networks. Andreas has a PhD in Deep Learning and Neural Systems from the University of Surrey.

Your apps made to order

Trusted by the world's leading brands

BBC logoMakro logoVirgin Unite logoNBC logoFujitsu logo
Your apps made to order

Related articles

120 mobile app ideas entrepreneurs MUST make in 2024

120 mobile app ideas entrepreneurs MUST make in 2024

Today’s the day to start the next unicorn 🦄

You’ve heard of superheroes but what is a Super App?

You’ve heard of superheroes but what is a Super App?

Super Apps create their own ecosystem with lots of functionalities.

141 procrastination-killing entrepreneur motivational quotes

141 procrastination-killing entrepreneur motivational quotes

“Opportunity does not knock. It presents itself when you beat down the door.” – Kyle Chandler

Liam Collins

Real success stories from real customers

See how companies like yours used’s app platform to help them achieve their business goals

How we helped the BBC run a world-class event experience

BBC Click producers needed an app that enabled their live audience to interact with polls and questions, which delivered in double-quick time.

How we saved Makro 98.3% on order management software costs

Asia’s largest cash and carry needed software that could scale with their rapid growth, so we built them something that allowed them to onboard new customers without technical hiccups.

How we helped Moodit’s users “crowdsource” positivity

With our help, Dr Hassan Yasin created a mental health app designed to help children and adolescents express their worries and improve their social connectedness.

Got questions?

Head to our FAQ page for in-depth answers

Read FAQs