Teaching Quantum Circuits to Learn Time: A Conversation with Dr. Moein Ivaki

Moein Ivaki, Aalto University Postdoc Researcher

1. Could you briefly introduce your postdoc project - what core question you’re tackling this year, and why it matters?

I’m a postdoc in the MSP group at Aalto University’s Department of Applied Physics. My project asks a simple but demanding question: which quantum systems are actually good learners for temporal tasks? I study how structural properties - nonlocal correlations, complexity, disorder, symmetries, noise - shape performance in quantum reservoir computing and temporal quantum machine learning. The aim is to link “physics knobs” to “learning capabilities,” so future quantum devices can be engineered not just to compute, but to learn patterns that are classically costly. More broadly, I’m interested in how quantum complexity relates to operational feasibility and learnability.

2. What methods or tools are central to your approach (experimental, computational, or theoretical), and how does this differ from how you worked during your PhD?

I combine large-scale numerical simulations with analytical modelling and simple toy models that can be mapped to realistic circuit architectures. We drive quantum circuits and many-body systems with time-dependent inputs and track both learning performance (prediction error, memory, generalization) and physical resources (entanglement, complexity, noise sensitivity). Compared with my PhD work on localization, criticality, and transport, the focus has shifted from understanding phases of matter to evaluating task performance under device-like constraints and machine-learning benchmarks.

3. Collaboration often shapes results: how have interactions with your PI, lab mates, or external partners influenced your direction or sharpened your thinking?

Discussions with my PI, Prof. Tapio Ala-Nissila, have been key to grounding abstract notions like “quantumness” or complexity in concrete, testable learning tasks. Regular meetings turn vague ideas into specific protocols we can simulate or suggest to experimentalists, which forces me to sharpen questions and avoid purely formal detours. Within the QRC consortium, colleagues with hardware, algorithmic, and data-driven perspectives constantly remind me to think in terms of what can be implemented, calibrated, and scaled. Collaborations with experimental and industry-adjacent partners also keep robustness and simplicity at the centre, not just theoretical optimality.

4. Tell us about a roadblock you’ve hit - technical, logistical, or conceptual - and what you changed to get past it.

A major conceptual roadblock was to quantify how much “quantumness” is actually useful before the architecture becomes too complex to scale. Early on, we compared very different models, making it hard to isolate genuine quantum effects from implementation details. Progress came when we built a tunable family of probabilistic circuits where complexity and entanglement can be varied systematically while the basic structure stays fixed. This let us probe learnability and scalability with a common set of tasks and metrics. Getting there meant simplifying models and letting go of elegant but impractical constructions; next we aim to extend this to noisy and symmetry-constrained systems.

5. If someone outside your field asked, “Who could use your findings in the next 3–5 years and how?”, what would you tell them?

Over the next 3–5 years, I see three main users. Quantum hardware teams can use our results as design guidance: which circuit structures and operating regimes are promising for temporal learning, and which are likely dead ends. Groups working on quantum-enhanced sensing and control could adapt our methods to build noise-aware predictors and filters that run directly on quantum platforms. On the theory side, our models provide clean benchmarks for new quantum learning ideas, helping distinguish genuine routes to advantage from hype, even before full-scale applications exist.

6. Looking ahead, which skills or perspectives from this postdoc are you most excited to carry forward, and what advice would you give PhD candidates eyeing a postdoc path?

This postdoc has pushed me to think end-to-end: from physical models, to algorithms, to what can actually be measured and controlled on a device. I’m keen to carry forward the ability to design minimal models that capture the essential physics of quantum learning, and to move fluently between ideas in condensed-matter, quantum information and complexity. For PhD candidates, I’d say: treat the postdoc as your first real chance to set your own direction. Choose a project and environment where you can deepen one core strength and genuinely add a new one - whether that’s coding, hardware, or a conceptual toolkit. Prestige matters less than finding a place where you can do sustained, curious work without burning out.