
Ana Carolina Cunha
Hi, I'm Ana. I'm an AI engineer at BMW, building language model systems over engineering data. The work I care about most is research, mostly on how these models fail and what their internals represent.
I studied Cognitive Science at UC San Diego, specializing in machine learning and neural networks, and I'm finishing a Master's in Informatics at TUM. Recent work has been on credit assignment in reasoning models, evaluation robustness for medical LLMs, and unsupervised structure in genome models.
Latest Research
Master–Novice Divergence for Reward Densification
Prompt Optimization in Medical LLM Benchmarks