We recently talked with Sopra Steria about its work and capabilities around big data and analytics. The company has created a Data Science CoE (or Technology Stream in Sopra Steria’s terminology), which brings specialized services and expertise to the various in-country systems integration (SI) business units, focusing on its key accounts and bringing vertical knowledge.

Sopra Steria’s Data Science unit has been developing AI-based use cases, focusing on the analysis of unstructured data, including through the use of computer vision technologies.

Applying AI: client examples

An example of its work in applying AI algorithms to unstructured data is an aircraft manufacturer client that Sopra Steria has been helping by developing solutions to automate the inspection of aircraft using pictures taken by drones. This approach, which is much faster than manual inspections, uses drones to take pictures which are stored and compared in real-time with a repository of pictures showing anomalies.

The process started with the drone and AI identifying simple items such as missing paint or screws on the aircraft, and it is now getting more complicated as Sopra Steria grows its expertise. Sopra Steria estimates that it requires ~ 200 pictures to teach the ML algorithm to spot anomalies, and believes its approach is now mature enough to be applied to similar projects with other clients.

Another example is a project based on the use of satellite images. Sopra Steria has helped an electricity grid operator to analyze its network and identify where it needs to prune trees, and prioritizing them. Unlike the aircraft example, this approach does not rely on edge-based computing, as flying drones in areas with many trees is a challenge. The broad principles are the same, however, and the approach helps prioritizing trees which are mostly likely to create interferences in the electricity grid.

Creating IP from expertise

Looking ahead, Sopra Steria’s Data Science unit wants to create IP out of its expertise. The CoE acknowledges it is walking a fine line between AI cloud vendors that tend to offer vertical-agnostic micro-services and ISVs that are high specialized (e.g. Sensetime and Ntechlab in China and Russia respectively, both around video surveillance). The unit is adopting two main approaches:

• A methodology approach for use case development. For example, in drone-based aircraft inspection, it knows what images of anomalies it needs to look at to optimize the learning process of the ML. And for email automation around language utterance, the unit created a repository of terms and jargon specific to the insurance industry
• An internal focus. The unit has been taking part in Sopra Steria’s Inner Source approach (see blog here) and is creating AI and ML micro-services that it wants its developers to use. Indeed, it is finding that its software developers have an appetite for using AI/ML micro-services it has created. The CoE is now acting as a support organization for using these micro-services and applying them to projects. We view this approach as a positive step in Sopra Steria’s evolving IP strategy. While Sopra Steria has been investing in commercial software products (e.g. Sopra HR Software, Sopra Banking Software and real estate), the firm’s SI units have been less vocal about their IP creation. This is now changing, initially driven by Inner Source to provide software developers with the software tools and environments they require. Sopra Steria is now accelerating its IP approach.