FELD M

Three published engagements are data science, and they are more specific about method than most anonymized work manages.

For a multinational healthcare corporation, document classification was built on deep-learning computer vision—CLIP for one-shot learning and image extraction, an XGBoost classifier as the benchmark—and the write-up compares its own results against AWS Comprehend, beating it on several labels. The pipeline runs in AWS and processes documents in the thousands.

For a media company, a propensity model on Random Forest and XGBoost turns Adobe Clickstream data into subscription predictions, running on Azure and written back into Adobe Analytics for user-level analysis. For an international retailer, a forecasting model in R and SparkR simulates price adjustments across more than 10,000 products, including substitution and cannibalization between them.

All three are anonymized by sector, while the named clients—Der Spiegel, Media Markt, SBB, Hugo Boss, bexio—sit on the analytics, tracking and consent side of the practice. That split is the thing to understand before engaging: this is a data and marketing-technology consultancy with a data-science team inside it, not an AI firm. Best for organizations whose AI ambition is downstream of a measurement problem.