Dan Alistarh

Dan Alistarh is a Professor at the Institute of Science and Technology Austria (IST Austria) and leads the Deep Algorithms and Systems Lab (DASLab). His research focuses on efficient algorithms and systems for machine learning, including distributed optimization, sparse and quantized neural networks, and scalable training/inference techniques. He holds a PhD from École Polytechnique Fédérale de Lausanne (EPFL) and has held positions at MIT, Microsoft Research, and ETH Zurich.

Research interests: Optimization for data analysis, parallel and distributed optimization, efficient machine learning algorithms, and distributed systems. Collaborations include work on optimization under uncertainty with Immanuel Bomze, Radu Bot, and others.

Publications span top venues like NeurIPS, ICML, and DISC, with notable contributions in model compression (e.g., GPTQ, SparseGPT), communication-efficient distributed training, and concurrency algorithms. Awards include ERC grants and best paper awards.

He advises a team of PhD students and postdocs, and his lab's tools are widely used (e.g., GitHub repositories). His work has been adopted in industry (e.g., OpenAI, Neural Magic).