AI for forecasting: foundations models, hype & best practice

Lightning Lessons

Practical skills & tools to accelerate your career

AI for forecasting: foundations models, hype & best practice

Hosted by Franz Kiraly

103 students

What you'll learn

forecasting foundation models - what they are, what they do

Gentle introduction to foundation and pre-trained models. Zero-shot, few-shot, fine-tuning; open vs gated, etc

best practice - handling software, weights, deps, evaluation

special considerations and "FMops" for forecasting - FM handling (best) practice, key difficulties & solutions

... and what the vendors don't want you to know about

lock-in tactics, influencer propaganda, gated/closed models, licenses, full-value open alternatives

Why this topic matters

Get most out of AI and FM for forecasting - learn about capabilities, pitfalls, best practice, how to deploy and reliably test cutting edge models on your own data. Learn how to navigate a landscape that has only benchmark winning models - full of gated subscriptions, and 100s of (absolutely independent) influencers promoting one gated model, conformal prediction, or AI startups without a product.

You'll learn from

Franz Kiraly

Director, German Center of Open Source AI, and founder of sktime

Franz Király is director of the German Center for Open Source AI. He has 9 years experience as university professor at top tier universities, and 10 years of experience in industrial data science and AI roles, including 5 years at the Alan Turing institute - all of which included mentoring and teaching responsibilities. Dr Király is also the original founder and a core developer of sktime, the leading package for forecasting and AI with time series.