Fluidly: Providing small businesses an intelligent financial service with Google Cloud Platform

About Fluidly

Fluidly helps business owners and their accountants sleep better at night by forecasting and optimizing their finances. At the forefront of intelligent cash flow, the company uses AI and data engineering to provide confidence, certainty, and control around cash flow to thousands of SMEs, working in partnership with leading accounting firms.

Industries: Technology, Financial Services
Location: United Kingdom

Fluidly migrated its financial platform to Google Cloud Platform for quicker deployment times, scalable machine learning and artificial intelligence capabilities, and reduced need for management.

Google Cloud Results

  • Reduces deployment times from 1 hour to under 5 minutes with Google Kubernetes Engine
  • Operates AI at scale, allowing the platform to service tens of thousands of clients with minimal management
  • Trains and deploys thousands of machine learning models on a platform streaming 600GB of data daily

Increases production releases from 4 a week to 36 a day

For small and medium-sized enterprises (SMEs), cash flow is king. Managing cash flow can be the key to survival but staying on top of it is difficult and detailed work, especially when resources are limited. Focused on the SME market, Fluidly has developed a platform that takes the hard work out of cash flow management and credit control. What makes it stand out from standard accounting packages is its unique combination of AI technology and financial modelling. With automated credit management tasks and intelligent cash flow forecasts, Fluidly gives companies a clear picture of their future cash position and helps them make the decisions that count.

"We're like an autopilot for your business," says Johnnie Ball, Chief Data Officer and Co-founder at Fluidly. "We keep customers flying in the right direction and flag up any potential future issues that they might have missed."

"Our previous infrastructure was fine when our product was simpler, but it couldn't support the support the more sophisticated AI and machine learning we wanted to use. We needed a way to realize our ambitions with AI and scale up with the business, and Google Cloud Platform has given us the solution."

Mike Hancock, CTO, Fluidly

Founded in 2016, Fluidly began as a small team of engineers and data scientists with an idea for a smart financial platform. After releasing an initial version, by early 2018 it was clear that the platform was a success and the company began plans for a wider commercial launch. By this point, however, Fluidly's initial infrastructure had reached its limits when combined with the requirements of new AI enhancements with the product. The existing infrastructure couldn't easily support the technology and expected growth the way Fluidly needed it to, so the company decided to make a change by migrating to Google Cloud Platform (GCP).

"Our previous infrastructure was fine when our product was simpler, but it couldn't support the more sophisticated AI and machine learning we wanted to use," says Mike Hancock, CTO at Fluidly. "We needed a way to realize our ambitions with AI and scale up with the business, and Google Cloud Platform has given us the solution."

AI functionality at scale with Google Cloud Platform

Fluidly's initial platform was built with an emphasis on speed and simplicity. As time went on, however, there was a clear need for stability and scalability. As the number of customers grew, services ran into problems and slowed down. The infrastructure's absence of monitoring meant that Fluidly couldn't identify and respond as quickly as it wanted to. Also, the lack of support for more complex AI features was a bottleneck on the development of the company's product. The cash flow forecasting feature, in particular, required building machine learning pipelines quickly and at a huge scale, which simply wasn't possible with the existing setup. It was time to look elsewhere.

In early 2018, after evaluating a range of cloud providers, Fluidly chose GCP because of its AI functionalities and ease of use. "We feel the tooling around machine learning and AI is head and shoulders above the competition with GCP," explains Johnnie. "It's great for us to be at the forefront of this kind of technology." As well as the depth of features, GCP has proven easier than other platforms to set up, configure, and maintain. This was crucial for Fluidly and its growing, dedicated team.

They soon got to work migrating to GCP, beginning with the machine learning elements and data pipelines. The first task of the new platform was to extract, aggregate, and compute the data before processing it with Cloud Dataflow and pipelining it to a range of different databases. Fluidly leverages the security of Cloud SQL to manage data, while the rest is in BigQuery or Cloud Memorystore, which allows Fluidly to use the open source REDIS protocol. The training data for the AI models is held in Cloud Storage before it is taken into Cloud Machine Learning Engine, the heart of Fluidly's platform.

"The simplicity of the service and its flexibility makes Google Cloud Machine Learning Engine absolutely key for us. With just a command, we can train a model, add our secret sauce, and then serve predictions through an API. I don't think we could do that with any other platform."

Johnnie Ball, Chief Data Officer and Co-founder, Fluidly

For the cash flow forecasting features to work, the platform has to train and deploy thousands of different models that take into account all kinds of variables and scenarios. Setting these up in Cloud Machine Learning Engine becomes very simple once the initial configurations have been made. Fluidly then introduces its own Python scripts to apply the models to the right set of data in its system, creating its unique forecasting solution.

"The simplicity of the service and its flexibility makes Google Cloud Machine Learning Engine absolutely key for us," says Johnnie. "With just a command, we can train a model, add our secret sauce, and then serve predictions through an API. I don't think we could do that with any other platform."

After successfully migrating its AI and machine learning components to GCP, Fluidly has begun to migrate the rest of its infrastructure over as well. The vast majority of its services run on Google Kubernetes Engine (GKE), providing ease of use and stability at scale. "With GKE, we get the power and simplicity of Kubernetes with none of the hassle of maintaining the clusters," explains Mike. To coordinate the system, Cloud Pub/Sub is used for messaging while Cloud Composer keeps things running on schedule. Given the sensitivity of the data it handles, Fluidly puts a premium on security and has built its own encryption management system with Cloud Key Management Service. Meanwhile, Google Data Studio integrates seamlessly with the databases for quick and painless business intelligence reporting.

"Since we started migrating to Google Cloud Platform in 2018, our customers have increased tenfold. But we're still using the same configurations we had last year, even at very high loads. It really helps us retain simplicity at scale."

Johnnie Ball, Chief Data Officer and Co-founder, Fluidly

Managed services for more freedom to experiment

With GCP, Fluidly has built an AI platform that is richer, easier to manage, faster, and more scalable than anything it had before. The fully featured cash flow forecasting and credit control features can now easily stream around 600 gigabytes of data a day, scaling up and down as needed. Existing services can be reproduced in minutes while new services can be designed, tested, and deployed in just three hours, compared to more than a day on the previous infrastructure. Fluidly develops and matures its features at a much faster rate now, with production releases accelerating from four a week to 36 a day. Meanwhile, data processing, which used to be done in overnight batches, now takes under five seconds. All this has helped keep the team moving with speed and keep the platform stable, as there has been a 900% increase in users within a year.

"Since we started migrating to Google Cloud Platform in 2018, our customers have increased tenfold. But we're still using the same configurations we had last year, even at very high loads," says Johnnie. "It really helps us retain simplicity at scale."

For Fluidly's team of data scientists, this simplicity and ease of use translates to a culture of experimentation and self-reliance. New ideas can be put into motion without having to worry about maintenance or scalability or waiting on engineers. This spirit has spread throughout the company, empowering teams to experiment quickly, iterate, and grow as the company expands.

Fluidly is focused on growing at an even greater rate this year and migrating the last remnants of its old infrastructure over to GCP. It has also started exploring new AI models with TensorFlow, the open source machine learning library developed at Google. "We're excited to use TensorFlow on GCP soon," says Johnnie. "It's the right place to deploy this kind of technology, and it's good to know we're now in the best position to get the most out of it."

About Fluidly

Fluidly helps business owners and their accountants sleep better at night by forecasting and optimizing their finances. At the forefront of intelligent cash flow, the company uses AI and data engineering to provide confidence, certainty, and control around cash flow to thousands of SMEs, working in partnership with leading accounting firms.

Industries: Technology, Financial Services
Location: United Kingdom
Google Cloud Platform logo

12 Months FREE TRIAL

Try Kubernetes Engine, BigQuery, and other Cloud Platform products with $300 in free credit and 12 months.

TRY IT FREE
Google Cloud Platform logo

12 Months FREE TRIAL

Try Kubernetes Engine, BigQuery, and other Cloud Platform products with $300 in free credit and 12 months.

TRY IT FREE