The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence) Productionizing Time-Series Workloads at Siemens Energy with Edgar Bahilo Rodriguez - #439
Dec 18, 2020
Edgar Bahilo Rodriguez, Lead Data Scientist at Siemens Energy, dives into the complexities of productionizing time-series workloads. He shares insights from his journey moving from energy engineering to machine learning, discussing the blend of technologies used to enhance forecasting models. Edgar highlights industrial applications in wind and energy management, emphasizing sustainable solutions. The conversation also touches on the evolution of machine learning in operations, automation in monitoring, and the integration of diverse programming languages in robust AI workflows.
AI Snips
Chapters
Transcript
Episode notes
Accidental Coder
- Edgar Bahilo Rodriguez, an energy engineer, accidentally learned to code while working on electricity price forecasting.
- He later transitioned to machine learning, focusing on time series forecasting at Siemens Energy.
One Model to Rule Them All
- Machine learning allows for diverse forecasting strategies, including direct forecasting and multiple input/output models.
- This "one model to rule them all" approach simplifies complex time series analysis.
Event-Driven Architecture
- Design machine learning systems with event-driven architectures, accommodating future scalability needs.
- Consider the trade-offs between numerous small models and a single large model.