About MILES

We founded our lab in 2023 at McMaster University as the Sustainable Systems and Methods Lab (SSM). The original mission of the lab was primed on various forms of sustainability, particularly technical (maintainability) and social/individual (human factors) sustainability. From day one, though, the lab has led cutting-edge research on model-driven engineering and machine learning, which grew into the primary (and almost exclusive) topic we focus on. To reflect this focus, we renamed the lab to MILES: the Model-Integrated and Learning-Enabled Systems Lab in 2026.

Mission

Our mission is to develop methods, techniques, and tools to position model-integrated engineering as a foundation for trustworthy learning-enabled systems.

Learning-enabled systems are increasingly powerful, but they are also difficult to engineer using traditional software development techniques alone. Their behaviour is often learned rather than explicitly programmed, making it harder to understand, constrain, verify, and assure.

At MILES, we investigate how models and learning can complement one another. Models provide explicit structure, semantics, constraints, and abstractions around learning-enabled components, while machine learning enables systems to acquire and adapt behaviors that would be difficult to specify manually.
We therefore study both directions of this relationship: models for learning, where model-driven techniques support the construction, execution, analysis, and assurance of learning-enabled systems; and learning for models, where machine learning supports activities such as model transformation, synthesis, adaptation, and evolution.

Our long-term vision is to make learning-enabled systems systematic to engineer, understandable in their behaviour, and amenable to rigorous verification and assurance.

Research goals

To advance this mission, our research focuses on the following goals.

• Engineer learning-enabled systems through explicit models
  We develop model-driven abstractions, metamodels, transformations, and architectures for representing learning problems, environments, system behavior, and interactions between learned and engineered components.
• Make learned behaviour easier to analyze and verify
  We investigate how explicit model-level representations can enable formal analysis, model checking, graph-based reasoning, testing, and other assurance techniques for systems whose behaviour is partly learned.
• Integrate models with learning at runtime
  We study runtime models, scene graphs, digital twins, and other causally connected representations that enable systems to monitor, reason about, and adapt their behavior during operation.
• Use machine learning to advance model-driven engineering
  We investigate how reinforcement learning, generative AI, and other learning methods can support model transformation, model construction, adaptation, synthesis, and software engineering tasks.
• Develop reusable foundations for model-integrated AI engineering
  We aim to identify reusable abstractions, architectures, transformations, patterns, and tools that allow learning-enabled systems to be engineered systematically rather than as isolated, one-off solutions.

Across these goals, a central question guides our work is: how can models bring engineering rigour to learning-enabled systems?

Research themes

Our work spans several closely connected research themes:

• Model-Driven Engineering for AI
  We use models, metamodels, and model transformations to structure the development of learning-enabled systems and make their behavior more explicit and analyzable.
• AI for Model-Driven Engineering
  We investigate how machine learning, reinforcement learning, and generative AI can support model transformation, synthesis, evolution, and other MDE activities.
• Verification and Assurance of Learning-Enabled Systems
  We study how model-level representations can support formal analysis, testing, verification, and assurance of systems whose behavior is learned or adaptive.
• Runtime Models and Digital Twins
  We develop model-based representations that remain connected to running systems and support monitoring, reasoning, adaptation, and decision-making at runtime.
• Reinforcement Learning and Adaptive Systems
  We investigate the engineering of RL-based systems, including learning environments, curricula, policies, human guidance, and adaptive behaviour.
• Graph-Based Representations and Reasoning
  We use graph-based models, including scene graphs and model transformation systems, to represent complex environments and support incremental reasoning and analysis.

Join MILES

MILES welcomes students interested in research at the intersection of software engineering, model-driven engineering, machine learning, and intelligent systems.
Our projects range from foundational research on models and learning to applied work involving reinforcement learning, digital twins, autonomous systems, runtime models, verification, and AI-assisted software engineering.

We regularly work with undergraduate, master’s, and PhD students.
Strong candidates are curious about both how intelligent systems learn and how such systems can be engineered rigorously. If you are interested in joining MILES, please get in touch with a short description of your background, research interests, and the topics you would be interested in exploring with us.

More info about the application process and open positions here: https://istvandavid.com/positions-and-application/.