Business challenge

Vision Banco realized that their existing x86 infrastructure they were deploying machine learning models on was not enough to handle the amount of data with the performance they needed. They needed GPU acceleration to handle the demands of AI and machine learning workloads.

Transformation

Vision Banco saved time and increased revenue by building more accurate models for credit risk analysis with H2O Driverless AI on IBM Power Systems AC922.

Results

Increased

number of credit products accepted per customer

Deployment

time reduced between two days and one week to two hours

Performed

more accurate cross-selling to increase revenue

Business challenge story

The Need for Increased Accuracy

Vision Banco was founded in 1992 and is based in Asuncion, Paraguay. They provide financial services to small and micro-sized companies in Paraguay. They also offer credit card services, remittances, utility and tax collection services, pension plan contribution plans and payment transfer services. The bank has 1,798 employees and manages over $997M USD in assets. 

The bank’s data scientists were challenged to expand the credit card services of existing customers, easily determine credit risks with better accuracy and better predict payment defaults. They have been a longtime IBM Cognos Analytics to find insights in their business data. Cognos Analytics is an AI-infused business-intelligence solutions to easily create compelling visualizations and dashboards.

Vision Banco realized that their existing x86 infrastructure they were deploying machine learning models on was not enough to handle the amount of data with the performance they needed. They needed GPU acceleration to handle the demands of AI and machine learning workloads.

IBM Business Partner Softshop helped guide Vision Banco to the right solution for their AI efforts. Since 1990, Softshop has been specializing in IT infrastructure and services for their clients. They have experience with enterprise AI and IBM Power Systems and knew the right platform to handle these workloads. Softshop introduced Vision Banco to H20 Driverless AI and IBM Power Systems AC922.

Transformation story

Developing More Accurate Machine Learning Models

Vision Banco deployed H2O Driverless AI on IBM Power Systems AC922 running Red Hat Enterprise Linux. IBM Power Systems AC922 are accelerated servers with NVIDIA Tesla V100 GPUs that deliver unprecedented performance for modern high performance computing, analytics and artificial intelligence. H2O Driverless AI employs the techniques of expert data scientists in an easy application to use that helps scale your data science and machine learning efforts. The combination of H2O Driverless AI on Power Systems AC922 is an easy on-ramp for machine learning and accelerates data ingest and training times for rapid development and deployment.     

The data science team at Vision Banco are responsible for machine learning and data analytics at the bank. “Building predictive models has never been so easy, in such a short time and with such good results" says Alejandro Lopez, team leader. “We manufactured our best models for collections on this platform. Collection management has completely improved recovery to levels never reached. It was easy to get started and was a seamless experience. Softshop came in and helped us deploy H20 Driverless AI and we were able to get started right away.”  

The bank is now migrating all of their models to H20 Driverless AI, as the team immediately saw the benefits of running their models on GPU accelerated servers. The performance and accuracy of their machine learning models immediately increased by deploying them onto H2O Driverless AI and Power Systems AC922. They are able to get insights into their data much faster than on traditional x86 infrastructure. The data science team and the rest of bank are now bringing valuable insights to the business that are improving customer service.

We are finally able to build predictive models with the least effort and time, making several new models in a single day, which used to take us months or weeks with other tools.

Alejandro Lopez, Data Scientist, Vision Banco

Results story

Saving Time and Increasing Revenue with Better Machine Learning

Since deploying H2O Driverless AI on IBM Power Systems AC922, the data science team has saved time and increased revenue by building and deploying models that have doubled the number of credit products accepted per customer. Additionally, the team was able to pinpoint credit and default risks with greater accuracy than previously.

“We are finally able to build predictive models with the least effort and time, making several new models in a single day, which used to take us months or weeks with other tools" says Alejandro. "Working on Driverless AI on IBM Power Systems AC922 has generated great expectations that so far have been translated into results. More than 90 % of credit loan request are being processed with the models we build. That is the trust we are placing on this platform."

Previously, it took almost a week or more than two days to deploy a model on their old x86 infrastructure. With H2O Driverless AI on Power Systems AC922, they can now deploy and run a model in less than two hours. This is critical to improving customer service by being able to derive insights from their data and make personalized offers in real-time. The team is also seeing benefits from their new AI platform across the entire business, such as in accounts payable. 

The bank is now looking at utilizing IBM Watson Machine Learning on their Power Systems AC922 accelerated servers. Watson Machine Learning Accelerator combines popular open source deep learning frameworks and efficient AI development tools to bring the benefits of AI into your business. They are starting to work on natural language processing with TensorFlow to create even more value for the bank from their data. Vision Banco will continue to gain competitive advantages and improve customer service with enterprise AI on IBM Power Systems AC922.

Working with Driverless AI on IBM Power Systems AC922, we have generated great expectations that so far have been translated into results, more than 90 % of credit loan requests are being processed with the models we build, that is the trust we are placing on this platform.

Alejandro Lopez, Data Scientist, Vision Banco

business logo

About Vision Banco

Vision Banco was founded in 1992 and is based in Asuncion, Paraguay. They provide financial services to small and micro-sized companies in Paraguay. They also offer credit card services, remittances, utility and tax collection services, pension plan contribution plans and payment transfer services. The bank has 1,798 employees and manages over $997M USD in assets. 

Take the next step

To learn more about IBM Power Systems AC922 servers, please contact your IBM representative or IBM Business Partner, or visit https://www.ibm.com/us-en/marketplace/power-systems-ac922

To learn more about H2O Driverless AI, please contact your IBM representative or IBM Business Partner, or visit https://www.h2o.ai/products/h2o-driverless-ai/