Visio.ai: Creating an innovative business model using Cloud TPU
About Visio.ai
Visio.ai creates managerial artificial intelligence for retail stores using computer vision and machine learning.
Tell us your challenge. We're here to help.
Contact usUsing the Google Cloud solution, Visio.ai managed to expand its product portfolio and scale the number of customers with healthy margins.
Google Cloud results
- 95% reduction in analysis time of customer operations
- +300% increase in the number of customers served
- 50% savings in inference costs
- Daily traffic volume equivalent to four months of images
- Cost reductions for end customers
6x higher performance enables new products
Visio.ai uses computer vision and artificial intelligence to create a virtual manager/copilot for retail stores. The purpose is to provide accurate, relevant information on store operation so that owners and managers can prioritize tasks daily to maximize operating efficiency and profitability, even if they are not around all the time.
Currently, the startup’s products enable store owners to reduce the time spent analyzing cashier fraud, monitoring daily operations, and checking their units’ operating hours.
Visio.ai’s journey was supported by Google Cloud at all times, even during change and growth. A significant chapter in this story was the company’s involvement in Google for Startups Accelerator, a program that aims to support successful, diverse, and inclusive startup communities around the globe. The team had access to mentorships and resources to test models and solutions before actually serving a higher number of customers.
These tests led to the most crucial turning points for the startup within Google Cloud’s own environment. Initially, Visio.ai had been using GPUs (graphics processing units) to process visual data, that is, images captured by cameras.
GPUs seemed to be the most appropriate architecture. However, it took only a detailed assessment to see a new finding: the solution’s processing volume did not match the scalability the company had hoped for. As GPUs have fixed capabilities, reaching a balance proved hard: the company either hired well above or well below what was needed. There were no linear costs.
In this quest for transformation and improvement, the Google for Startups team pointed out a more feasible alternative: Cloud TPU (tensor processing unit), a specific machine-learning workload solution.
Turning point and technological advancement with TPUs
With Cloud TPU, companies can accelerate machine-learning models, including natural language processing, classification, recommendations, and computer vision. Supported by Google Cloud, Visio.ai began making the changes it had longed for.
The team performed tests and continued unveiling everything the solution had to offer. Today, the company collects data from cameras and sends them to Google Cloud, processes them with Cloud TPU, and displays that information on its own platform. All of this with the security the cloud guarantees.
"Google Cloud had our back from day one. We were one of the first to use Cloud TPU and that’s how we managed to be one step ahead in terms of artificial intelligence. Using Google Cloud’s innovative tools and infrastructure, we managed to move forward and be at the frontier of what’s possible to produce with AI."
—Matheus Flores, Cofounder and CEO of Visio.aiThe architecture behind TPUs, divided into pods with several interconnected units, provided greater flexibility to distribute processing loads and scale as needed. As a result, Visio.ai managed to optimize processing costs and, consequently, customer costs.
"Not only does Cloud TPU offer operational support, it also has a strong impact on costs and scalability. This is the key differentiator that makes products economically feasible and generates ROI for end customers."
—Matheus Flores, Cofounder and CEO of Visio.aiResults that ensure growth
The advantages secured with the progressive use of Cloud TPU are crystal clear for the company: scalability and service delivery speeds combine with a 50% reduction in inference costs and a 6 times faster processing performance.
These benefits couple with an infrastructure use capacity that has barely reached 15%, which leaves room for tremendous growth in the future. In total, approximately four months of images are processed daily by Google Cloud. So far, over a million transactions have already been analyzed, which resulted in a 95% decrease in the time customers spend analyzing their own operation.
And speaking of victories, in 2022, Visio.ai expanded the number of stores served by over 300%. The determining factor was an increase in the capacity to serve new customers, driven by Cloud TPU’s scalability. The excellent results secured have caught the attention of investors and researchers from renowned universities in connection with the business and technology the startup deploys.
All the good news could only culminate in the company’s determination to further expand the project, as well as the possibilities offered by cloud tools. The team plans to dive into other areas within the realm of artificial intelligence and use Big Data along the way.
"We’re scaling fast and we have several choices ahead of us with Google Cloud. The future is not a leap you take, but rather an evolving curve. We’ve achieved this organic construction with orderly, strategic planning, and that will be our focus from here on out."
—Matheus Flores, Cofounder and CEO of Visio.aiTell us your challenge. We're here to help.
Contact us
About Visio.ai
Visio.ai creates managerial artificial intelligence for retail stores using computer vision and machine learning.