Munich / Berkeley

Tiny networks. Recursive reasoning.

Deep learning with few parameters: I reimplement architectures that apply the same network multiple times.

Recursion with shared weightsAn input x goes into a network f with weights theta. The output h is fed back into the same network T times, followed by the final output.xŷfshared θh ← f(x, h) ×T3 layers → 1 layer ×T
One network, applied T times.

At a glance

Studies
Management & Technology, TU München
Exchange
UC Berkeley, 2026/27
Currently
Reimplementing TRM in Keras
Specialization
Computer Engineering

Work

Projects

All projects →

Contact

Write to me.

No form, no detour. An email reaches me directly.