Deepomatic expands its AI vision with real-time data streaming

Challenge

Monolithic architecture slowed development

Deepomatic is an AI-powered computer vision platform for multiple image recognition and analysis at enterprise scale, providing game-changing visibility into field operations to automate quality control, optimize costs, and enhance productivity.

With a strong presence in telecommunications, utilities, and green energy companies, Deepomatic analyzes 20 million field operations annually, taking less than two seconds to analyze a photo.

Speed and flexibility are non-negotiables for Deepomatic. However, the team struggled with the rigidity of their monolithic services and used a complex API to access data.

“We were facing multiple problems. It was hard to build new services and slow to build new features because it was difficult to access data.” — Juliette Tisseyre, VP of Engineering, Deepomatic.

So, Deepomatic set out to find a data streaming solution that would help them transition to microservices, break up the APIs, uncouple their web app and databases, and improve the scalability of their architecture.

Why Redpanda

A simplified, scalable architecture and fully managed services

Choosing a managed solution was crucial to Deepomatic as it would free the team to focus solely on building their product. Not to mention the cost-savings from only needing two DevOps specialists to maintain the whole cloud infrastructure.

“We chose Redpanda for two main reasons: we wanted a fully managed service and we wanted the best.”

With Redpanda’s simplified streaming data platform, Deepomatic’s engineers could break down their monolithic services and separate their front/read APIs from the operational ones — resulting in a more flexible and scalable architecture that was infinitely easier to maintain and deploy.

Plus, Redpanda’s famously friendly support team made it a seamless switch, with Juliette describing the dynamic as more of a partnership than a vendor-customer relationship.

“It was quite easy for the first user to integrate Redpanda, and it was simple to onboard engineers as we introduced it to the whole team,” Juliette says. “We have software engineers, solutions engineers, and DevOps engineers coming onto the system, and everyone who has worked with Redpanda is very happy.”

Results

12x faster feature shipping and the freedom to innovate at scale

With Redpanda in its corner, Deepomatic has a stable, reliable, and flexible infrastructure to streamline its customers' business processes. By removing the roadblocks when building custom enterprise features, Deepomatic can optimize its services and has already cut feature development time from six months to just two weeks.

Today, 50% of their customer’s business logic operations rely on data streams. These operations include real-time data sharing and alerting the customers if something needs human review (like fraud or a suspected equipment issue).

"Every time we ship a new feature or microservice, we can do it much quicker with Redpanda. We can be far more agile and flexible, and respond to demands for new features within days." 

With Redpanda as their single source of truth, teams can quickly access the data they need, build data pipelines, and ship cutting-edge features to meet their customers' needs much sooner.

As a real-world example, in the use case of fiber optic activation Deepomatic typically performs real-time photo analysis using AI to check for equipment degradation. This analysis would take up to 20 seconds, keeping field technicians waiting for the results before they can start working. Developing an async solution would’ve taken months, but with Redpanda, a solution engineer built a stream-based microservice in just a few days.

Now, field technicians can get the data they need within single-digit seconds, while the back office gets the full assessment in real time. Improved efficiency for Deeopmatic — streamlined services for the customer. Win-win.