+--------------+
Honglu Fan  |
+--------------+


Who am I

I am both a pure mathematician and an ML practitioner.

Pure Mathematics

  • Scientifique Collaborateur II at University of Geneva
    • Principal Investigator of Ambizione PZ00P2-193324 by Swiss National Science Foundation
  • Postdoc @ Université d’Angers
  • Postdoc @ ETH Zürich

Specialize in algebraic geometry. My PhD + Postdoc has been an amazing decade of extended college life working on the frontier of mathematics.

Publications

Earlier papers (2015–2018)

2018

2017

2016

2015

  • Quantum Leray-Hirsch for toric bundles. Studies quantum D-modules and GKZ systems of toric fibrations. The referee did not want to spend time helping me improve it, so I chose not to publish — the note is available upon request.

Invited talks

Conference & seminar talks (2016–2021)

I don’t record every seminar talk; conference talks tracked since 2018.

2021 — Aug 19–20, introductory Zoom lecture series (4 talks), Chungnam National University.

2019 — Sep 10, Zürich–Beijing Moduli workshop, PKU, Beijing.

2018 — Aug 23, Workshop on Higher Genus Gromov–Witten Invariants, ZJU, Hangzhou.

2017

  • Dec 28, Winter Young Algebraic Geometer Workshop, SUSTC, Shenzhen.
  • Oct 27, Algebraic geometry and moduli seminar, ETH Zürich.
  • May 5, Algebraic geometry and moduli seminar, ETH Zürich.

2016

  • Nov 28, Geometry and Physics seminar, University of Michigan, Ann Arbor.
  • Oct 18, Algebraic Geometry seminar, University of California, Santa Barbara.
  • Sep 15, FRAGMENT seminar, Colorado State University.
  • Apr 10, AMS Spring Western Sectional Meeting (Special Session), University of Utah.

Teaching

Courses taught (2013–2019)

All teaching duties above are full lectures throughout the semester (including the PhD period).

Machine Learning

  • 2025 May - 2026 Jul, Research Engineer at Google DeepMind
  • 2023 Dec - 2025 Apr, Founding Engineer at poolside.ai
  • 2021 Oct - 2023 Dec, Research staff at Eleuther AI

Publications

Acknowledgments

Eleuther AI is what made all of these above possible at all. In addition, a lot of thanks need to go to Hi-Flyer AI which allowed me to abuse their 10,000-A100 cluster. Also need to thank Google Research Credit and Google TRC project for allowing me to use their amazing VMs and TPU pods.

Contact