Key results
The company
Puma Energy
pumaenergy.comGlobal energy company providing fuel distribution, retail, and aviation services.
Result highlights
- Data latency cut from 5 hours to 5 minutes
- Safety audit prep time cut from 2 weeks to 1 hour
- 20%+ reduction in fuel stockouts
The story
A global downstream energy retailer supplies fuel, lubricants, and aviation products to retail stations and B2B customers across Latin America, Africa, and Asia-Pacific.
Data fragmentation and slow dashboards forced staff to rely on manual workarounds, with 95% of usage involving raw data exports. Insights lagged by up to five hours, while safety and audit preparation required two weeks of manual effort.
The company consolidated data onto Databricks, using Delta Lake for storage and Unity Catalog to govern access across regions. Teams deployed retrieval-augmented generation (RAG) to synthesize unstructured feedback from social media and support tickets, turning disconnected text into actionable summaries. An automated MLOps pipeline manages a mix of models including GPT and Llama to serve diverse use cases, from predicting fuel stockouts to generating compliance reports.
Quotes
“Databricks has become the backbone of how we manage and use data. Since our teams can move faster, work smarter and deliver better experiences to the communities we serve, we’ve reduced fuel stockouts and scaled new capabilities across 15+ countries.”