My long-term goal is to make AI more responsible—accessible, sustainable, and righteous.
Currently, I am focusing on various facets of Efficient ML. Together with EffLers, I am striving to enable ML without giant-scale models/data/compute by innovating relevant theories, algorithms, and systems. Some recent research questions that I am specifically interested in are:
- How (and to what extent) can we reduce the data-level redundancy for efficient training?
- How can we replace the noise inherent in ML with a compute-efficient one?
Publications. See the group webpage.