Research topics:
- Alignment
- Large language models
- Natural language processing
- Mechanistic interpretability
Applications are invited for a postdoctoral position at EPFL, to be hosted by the Data Science & AI Lab, headed by Prof. Robert West.
We seek a candidate who will lead ground-breaking research projects with the goal of building safe AI that truly cares about human interests.
Instead of viewing alignment as mere postprocessing of pretrained models (“lipstick-on-a-pig alignment”), you will contribute to our agenda of “raising” models that will be continuously aligned throughout the training process, starting from token 1 in pretraining (see here for a high-level vision). To this end, you will combine synthetic data generation, mechanistic interpretability, LLM pre- and posttraining, and model evaluation in novel ways.
