I am a deep learning engineer at NVIDIA, focusing on frontier model co-design for efficient and agentic inference. Previously, I was at Microsoft Research. I graduated from the University of Michigan in May 2022 with a PhD in computer science working on graph + text learning with Transformers.
See Google Scholar for an up-to-date publications list.
Contact: tarasafavi [at] microsoft.com
Recent news
- November 2022: Looking forward to giving virtual keynotes on graphs + LMs at the MLoG and KG workshops at ICDM 2022. Thank you to the organizers for the invitations!
- November 2022: Really honored to receive the Kuck Dissertation Prize from UMich CSE for my dissertation! Thank you!
- November 2022: Our paper combining graphs and language models with CascadER will be presented orally at AKBC 2022! Joint work with my awesome collaborators Doug and Tom.
- July 2022: Started at MSR Redmond.
- July 2022: Got an ICML Outstanding Reviewer Award!
- May 2022: Successfully defended my PhD thesis! Thank you to my committee Danai, Paul, Kevyn, and Lu.
- February 2022: Had a wonderful time virtually visiting the Samsung AI Center Toronto and giving an invited talk on graph + text learning. Thank you for the invitation!
Misc (updated periodically)
- My advice on applying for the NSF GRFP, last updated in 2020
- All-time favorite books/audiobooks
- Middlemarch by George Eliot - narrated by the incomparable Juliet Stevenson
- The Master and Margarita by Mikhail Bulgakov - narrated by Julian Rhind-Tutt
- East of Eden by John Steinbeck - narrated by Richard Poe
- Project Hail Mary by Andy Weir - narrated by Ray Porter
- Vanity Fair by William Makepeace Thackeray - narrated by Frederick Davidson
- The Remains of the Day by Kazuo Ishiguro - narrated by Nicholas Guy Smith
- The Bonfire of the Vanities by Tom Wolfe
- The Orphan Master’s Son by Adam Johnson
- War and Peace by Leo Tolstoy - Pevear and Volokhonsky English translation