2026-04-22 All-hands meeting notes

Project Updates

JE

• Lily Wang is on an extended vacation and Chapin Cavendar is on parental leave, so some science projects may be noticeably slower

• Most of the team will be in Barcelona for meetings the week of May 4, 2026

• James and Ashley will be online

• https://docs.google.com/presentation/d/1zGPQYkedSvgh7t0hT9rNtiOOiAqf5YkTGxuQyr_1_CY/

•  

•  

•  

Science Updates

Recorded

https://youtu.be/ZIw2JlCfSso
https://docs.google.com/presentation/d/1e8dBEdxmDhxQ9bcT59nsA2XnKB8mlcJwRc0b0soPJDE/edit?usp=sharing

• MS – is fitting with PyTorch the same as fitting with smee?

• JW – exactly.

Infrastructure Updates

JW

https://youtu.be/A19tLhs2a6s

https://docs.google.com/presentation/d/1YhLPAiiwgQXxB1sIn23q4o7ghkycaCy9zBqWx17H92Q/edit?usp=sharing

Shouts-outs

• JW – shout-out to Matt and Ashley for picking up this science workload.

• JC – Shout out PB for putting together the N, P, and S datasets. That’s been on our radar for a long time but we never had the time to put it together.

•  

Q&A

• MS – I’m soliciting wish-list ideas for long-term infrastructure investment for a grant proposal I’m working on. If there’s something you’ve dreamt of that would take more than a year to implement, let me know!

• …

• JE – Surrogate modeling workflow could be a good one in this topic/area.

  • MS – Something like gaussian process model informed by condensed phase sims.

• JE – Also NAGL rebuild to avoid DGL

  • MS – I think that fits well under charge support framework.

• MS – Making it easier to test functional forms. But this is already kinda supported in openff-plugins. Eg. someone wants to try new mixing rules/LJ terms. To some extent this works already (eg. DCole’s team did DEXP and MGilson’s group did direct pol) but

  • JW –

  • MT – I feel strange about asking how to make things usable, when I’ve already made things usable. The added details of

  • MS – Eg testing new funcitonal forms in YAMMBS may be possible, but extending to ligand binding free energies is hard.

  • MT – One possible outcome of that is that we don’t provide additional things here, but make it straightforward for people to use our infra for part of a fitting workflow and their own components for others.

  • MS – I’ll start a draft here and circulate for feedback.

  • PB (chat) – I think it was easy to work with as it is now, my undergrad was able to add new improper functional forms and test on her own fork

  •  

  •  

•  

•