SKF customer-facing GenAI assistant

Problem

Customers needed faster answers to L1/L2 technical questions grounded in SKF product knowledge.

Contribution

Led delivery of a multi-agent assistant using open SKF data and retrieval. Scope included ingestion, static and dynamic RAG, agent/tool routing, monitoring, evaluation, and feature delivery into production.

Design decision

The system combined retrieval and agent/tool routing with monitoring and evaluation, rather than presenting an ungrounded general-purpose chat interface.

Outcome and evidence

Earlier CV records report a 28-second average response time and more than 100,000 external queries per month. These historical figures have not been independently revalidated for this page; baselines, periods, and attribution should be confirmed before reuse.

Source: public CV online or dated CV (PDF). No private project documents are linked.