projects

Award · Overcoming language barriers in healthcare with AI-powered reporting.
Team: Enough Slices
🥈 2nd Place at TUM.ai Makeathon (April 2023).

Caire is an AI-powered assistant designed to automate medical reporting. Using real-time speech-to-text and a fine-tuned LLM, it extracts medically relevant information and generates structured reports.

Publication · Black-box alignment for diffusion and flow models.
Niklas Schweiger, K. Ram, Daniel Cremers
Accepted @ ReALM-GEN Workshop, ICLR 2026.
📄 Download Paper

We propose a simple trust-region based search algorithm (TRS) which treats the pre-trained generative and reward models as black boxes, only optimizing the source noise to achieve reward-agnostic alignment.