About the Author
Every part of this book started as something I had to solve for real, not as a topic I picked to write about.
The earliest was disaggregation: teaching a model to look at one aggregate power reading and say which appliance was drawing it, the same problem Part 2 of this book covers. That question led to the one utilities actually pay for: not just what’s running now, but what demand and solar output will look like tomorrow, with an honest estimate of how wrong that forecast might be. My PhD took that question to the low-voltage substation, forecasting net load in the presence of solar and EV charging when a headline accuracy number wasn’t good enough to act on. At Eaton’s Centre for Intelligent Power, I ran that same question as Lead Data Scientist and Principal Investigator, securing €1.6M in competitive research funding and picking up the Eaton STAR Leadership Award along the way.
Today, as Principal ML Engineer at Analog Devices, the question is the same one this book keeps returning to: not whether a model works in a notebook, but whether it survives contact with production, supply chain demand planning and edge inference in my case. That is the bar every chapter here holds itself to.
The open-source side of that decade is public: Twiga (point and probabilistic forecasting, one interface, every model), MLPForecast (uncertainty-aware net-load forecasting), and Deep-NILMtk (a community toolkit for non-intrusive load monitoring, the same disaggregation problem this book opens with).
Projects → sambaiga.github.io/pages/projects
The published record runs the same thread: non-intrusive load monitoring, probabilistic forecasting, uncertainty quantification, physics-based intelligence, and AI for reliability engineering, across 10+ peer-reviewed papers in IEEE Transactions on Smart Grid, IEEE Transactions on Power Systems, and IEEE Transactions on Industrial Informatics.
Publications → sambaiga.github.io/pages/publication · ORCID
I also write about machine learning, Bayesian thinking, and applied AI at: