Tutorial on
Automated Machine Learning 

Automated machine learning (AutoML) has matured from a niche research interest into a cornerstone of modern AI development. By automating the selection, composition, and configuration of machine learning pipelines, including hyperparameter optimization, neural architecture search (NAS), and meta-learning, AutoML dramatically reduces the expert knowledge and time required to build high-performing AI systems. Over the past decade, the field has produced influential open-source tools (e.g., Auto-sklearn, SMAC, BOHB, DARTS, AutoGluon, TabPFN), attracted major industrial investment, and demonstrated competitive or state-of-the-art results across tabular, image, time-series, text, and scientific data. Yet, the landscape is shifting rapidly: the rise of large foundation models, growing demands for transparency and fairness, acute concerns about energy consumption, and the need for deployment on resource-constrained edge devices are all reshaping what AutoML means and how it must evolve. At the same time, new agentic approaches, such as AI Scientist or Darwin-Gödel machines, offer a completely new flavor of AutoML, opening up new opportunities to automate ML research.

This tutorial provides a structured, half-day deep-dive into both the well-established pillars of AutoML and the emerging frontiers that are redefining the field. We begin with a concise historical and conceptual overview, then move through six thematic blocks: (1) the synergies and tensions between AutoML and foundation models (including LLM-based auto-prompting, agentic AutoML, and tabular foundation models); (2) human-centered AutoML emphasizing interactivity, explainability, and fairness; (3) neural architecture search:

its principles, hardware-aware and federated variants, and its intersection with foundation model fine-tuning; (4) AutoML for resource-constrained environments, covering automated deployment, model compression, and green AutoML; and (5) a forward-looking discussion of open challenges, including benchmarking, scientific discovery, and sustainability in the era of large models.

Presenters

Schedule

This tutorial will be held in-person at the International Joint Conference on AI - European Conference on AI, Bremen, Germany, August 15-19 2026.

This half-day tutorial offers a comprehensive, up-to-date tour of automated machine learning (AutoML), spanning its classical foundations, its integration with modern foundation models and LLMs, human-centered design principles, neural architecture search, resource/sustainability-aware methods, and an outlook to the most recent state-of-the-art approaches. Attendees will leave equipped with both the theoretical grounding and practical intuitions needed to apply and advance AutoML in research and industry settings.

The outline will be as follows.

Outline

1. Introduction & The Evolving AutoML Landscape

2. Foundation Models & AutoML: Synergies & Challenges

3. Human-Centered AutoML

— Break —

4. Neural Architecture Search: Principles & Frontiers 

5. AutoML for Resource-Constrained Environments

6. Open Challenges & Future Directions

Target Audience

The tutorial targets AI researchers, machine learning practitioners, and graduate students who wish to gain a thorough, up-to-date understanding of the opportunities in AutoML. Attendees should have a solid grounding in basic machine learning concepts and familiarity with standard deep learning frameworks. No prior expertise in AutoML, hyperparameter optimization, or NAS is assumed. The tutorial is structured to be equally valuable to newcomers seeking a structured entry point and to experienced practitioners looking to broaden their knowledge of recent advances, particularly at the intersection of AutoML with foundation models, human-centered design, and sustainability.

Slides

TBD