Research Interests

I am interested in how the brain builds reliable representations of the world during development, and in particular in the part that the cortex’s own internal structure plays in this process.

How the cortex prepares for seeing before it sees

Before eye opening, the visual cortex already shows richly structured spontaneous activity. I want to understand what this endogenous structure is good for: how it is shaped by recurrent connectivity, and how far it constrains what the cortex can later learn from experience.

How experience turns variable responses into reliable ones

At eye opening, responses to the same stimulus differ from trial to trial; within days they become reliable. I am interested in the circuit mechanisms behind this transition. Our work so far suggests that reliability emerges when visual input becomes aligned with the pre-existing network structure, rather than through a general reduction of noise. See Trägenap et al., Nature Neuroscience 2025 (press release · free-to-read version).

Feedforward input meets recurrent circuits

How do feedforward inputs and recurrent cortical circuits shape each other as the cortex matures? With the Fitzpatrick lab we found that coherent cortical responses develop as feedforward inputs become more discriminable and better aligned with the recurrent network. See Lempel, Trägenap et al., Neuron 2025 (press release · doi:10.1016/j.neuron.2025.08.014).

Selective amplification across the cortical sheet

Which activity patterns does the developing cortex amplify, and which does it suppress? With the Smith lab we study selective amplification on the scale of millimetres and ask which recurrent subnetworks can account for it. See Mulholland, Trägenap et al., bioRxiv 2026 (co-first authors; doi:10.64898/2026.03.31.715687).

Looking ahead: across the animal tree

Most of what we know about cortical development comes from a handful of species. Yet brains differ, shaped by different evolutionary histories and different ways of seeing the world. By comparing how the cortex is organised in ferrets, cats, tree shrews and mice, I hope to tell apart the developmental principles that are shared across species from those that are particular to one. Where a species departs from the others, the departure itself points to the mechanism at work.

Looking ahead: from biology to learning machines

The developing cortex solves a problem that artificial systems also face: how to learn quickly from limited experience. I am curious whether the strategy we see in the brain, starting from a structured prior and aligning input to it, could make learning in artificial neural networks faster or more robust.

Approach

I work at the interface of theory and experiment: dynamical recurrent network models, analysis of large-scale calcium imaging data, and the comparison of model predictions with chronically recorded neural populations. I enjoy working closely with experimental partners and being part of the loop between theory and experiment that drives science forward: a model can suggest what to measure, and the measurement reshapes the model. In my PhD work I have collaborated with the labs of David Fitzpatrick (Max Planck Florida Institute for Neuroscience) and Gordon Smith (University of Minnesota).

Recorded talks

  • Experience drives the development of novel, reliable cortical sensory representations from endogenous networks. Bernstein Conference 2022, Berlin. Video

Code and data