Data Science and Machine Learning applied to your operations

Suez guarantees its client communities a daily calculated network efficiency, even though consumption readings arrive incomplete, due to remote meter reading failures and unequipped meters. Galadrim developed the estimation engine that fills these gaps: it predicts the consumption of meters with historical data, taking into account the weather and neighbouring consumption, reconstructs a typical profile for the least equipped parks, and delivers overall efficiency with its margin of uncertainty. The engine runs on Suez's Google Cloud infrastructure.

Castrol, a subsidiary of the BP group, operates an automated filling line where entering during the cycle exposes one to a serious accident, on an entirely offline site. We trained a detection model using several hours of video and 500 annotated images, chose the equipment capable of running it on-site, then installed the system on the line. A camera continuously monitors the dangerous zone: as soon as a person enters it, the alert goes off immediately, and no image leaves the factory.

Chanel's neuroscience division sought to understand the actual use of its products from filmed tests, with the aim of obtaining quantifiable data on how a person applies a cosmetic. Galadrim developed image recognition algorithms that track the user's head, hand, and fingers throughout the video. Each gesture is counted and classified by nature, whether a semi-circular massage or a tapping motion, which gives Chanel an objective measure of how a cosmetic is applied.

Vinci Construction daily assigns agents, site machinery, and locomotives to maintain the continuity of its railway operations. The rules to comply with are accumulating: rest periods, an agent's authorisation to operate a given piece of machinery, the sequence of journeys from one site to the next. Galadrim developed the planning engine, integrated into its internal tool, which processes all sites in a single calculation. A site manager opens their tool in the morning and finds their team's schedule already built and compliant with all rules, which they then only need to validate.

Groupe Blondel handles the logistics for the Chantiers de l'Atlantique site, where cranes, lifts, and hoists dictate the progress of the entire site. The challenge was to keep them operating at full capacity while assigning teams and tracking parcel flows, without creating queues in front of an overloaded resource. Galadrim developed the application that combines these constraints: a model predicts the incoming parcel volume to anticipate the real load, project managers reserve their lifting slots from the interface, and the assignment algorithm builds schedules respecting working hours and compatibilities between operators and machines. Each employee receives their personalised schedule in the morning.

Spoticar, Stellantis group's used vehicle brand, manages a network of resellers and the classifieds portal that brings them their buyers. The information crucial for a purchase was spread across six non-communicating systems: stock, maintenance, manufacturer documentation, customer reviews, and search histories. Galadrim brought them together into the group's single database, working within its ingestion framework and security rules, then built the dictionary that describes this data using business vocabulary. A buyer asks their question in French and gets the vehicles that match, a dealer manages their stock, all based on the same data.