Stipendiatin
Bahar Hamzehei

Dein Name: Bahar Hamzehei

Universität: HAW

Fachliche Rahmung Deines Promotionsprojektes:
Research School, Doctoral Programme in Sustainable Technologies and Systems (STS) at HAW Hamburg. My doctoral research focuses on reliable model-informed inference for partially observed dynamical systems, with applications in biomedical and other engineering domains.

Welche drei Hashtags beschreiben Deine Fachexpertise am besten?

#PhysicsInformedMachineLearning #DynamicalSystems #ReliableAI

Wie würdest Du fachfremden Kolleg:innen Dein Promotionsprojekt beschreiben? 

In many real-world systems, we can observe only a small part of what is actually happening. In intensive care, for example, clinicians have measurements such as heart rate, blood pressure or laboratory values, but these provide only an incomplete picture of a patient’s underlying condition. My PhD investigates how machine learning can be combined with mathematical models of system dynamics to infer what cannot be observed directly, quantify how uncertain those inferences are, and recognise when a prediction should not be trusted. I will first validate these methods on controlled reference systems and then study their use in biomedical data, with sepsis as a main case study, as well as in other engineering applications.

Von welchen Deiner Veröffentlichungen, Ausstellungen o. Ä. sollten andere unbedingt wissen?

My master’s thesis at the University of Bologna, “Decoding Spatial Pain Anticipation from EEG–TMS Data Using Deep Learning,” investigated whether neural signals could distinguish anticipation of left-sided pain, right-sided pain, or no stimulation. I developed the full analysis pipeline, from EEG preprocessing and quality control to time–frequency analysis and subject-wise model evaluation, and compared several deep-learning approaches including TCNs, Attention-TCNs, spectrogram-based CNNs and Transformer prototypes. I also examined alpha- and theta-band activity to investigate neural patterns related to threat anticipation and spatial processing, with particular attention to whether the models generalised to unseen participants. Related research was presented at NeuroMI 2025 in Milan and DOOCN 2026 in Berlin.

Wie bist du auf Pro Exzellenzia lead aufmerksam geworden und worauf freust du dich in deiner Stipendiatinnenzeit besonders?

I first learned about Pro Exzellenzia lead through my doctoral supervisor, Prof. Wolfgang Renz. I am particularly looking forward to the individual coaching and to exchanging experiences with other women in research. At the start of my doctorate, I would like to develop a clearer long-term career direction, strengthen my leadership skills and build a supportive professional network in Hamburg.

Was sollten andere von Dir als professioneller Person noch unbedingt wissen?

My background in industrial engineering still shapes the way I approach AI research: I am interested not only in prediction, but in understanding how a system behaves and what information can realistically be recovered from data. During my master’s work with EEG–TMS recordings, I became particularly interested in problems where signals are noisy, observations are incomplete and a model may perform well overall while still failing to generalise to new people. That perspective now carries into my doctoral work, where I focus on methods that are not only predictive, but also explicit about their uncertainty and limitations.

Was meinst Du, wo werden wir Dich beruflich in 5 Jahren antreffen?

In five years, I would like to be working as a researcher at the intersection of AI, dynamical systems and biomedical applications, ideally leading or co-leading interdisciplinary projects. I want to develop methods that go beyond benchmark performance and can be evaluated in terms of robustness, uncertainty and real-world usefulness. I also hope to be supervising younger researchers and building collaborations between people who develop new computational methods and those who work with complex real-world systems.