Postdoctoral Appointee - Federated Learning and Foundation Model

A postdoctoral appointee in Argonne National Laboratory's Mathematics and Computer Science Division works on the methods underneath training foundation models on scientific data: the mathematics of stochastic gradient training dynamics, and algorithms that make training efficient, reliable and private across distributed and federated settings, from theory through to open-source software. The appointee designs, implements and benchmarks new training algorithms, contributes research software, leads papers at leading machine learning and optimization venues, collaborates with universities and national laboratories, and has access to some of the world's most powerful computing systems.

The expected hiring range is $72,879–$121,465.

Ph.D. (within the last 0–5 years) with a publication record in optimization for machine learning, deep learning theory, or distributed or federated learning; strong Python and PyTorch skills including GPU training; coursework or research in optimization, probability or machine learning. Desired: stochastic optimization or loss-landscape analysis, differential privacy, fine-tuning or pretraining foundation models (e.g. LoRA), multi-node distributed training (DeepSpeed, FSDP, MPI) and open-source contributions.