Martin Seifrid

Robust learning from literature data: Model generalizability and uncertainty for predicting conjugated polymer solution conformation Predicting solution conformation and aggregation of conjugated polymers remains a bottleneck for translating solution processing into controlled film microstruc

Officially official! The group's first-ever publication is online 🥳

#ChemSky #MatSky

doi.org/10.1063/5.03...

This AI has chemical expertise — and helps synthesize 35 new compounds An open-source program helps researchers bypass a major bottleneck in the process chemical synthesis.

In my @nature.com story, I explain how chemists at Yale have created a chemistry AI tool by fine-turning the LLAMA LLM to create 2,498 expert models. Their system can judge which of the different experts to refer a query about reaction conditions to. Impressive! 🧪

www.nature.com/articles/d41...

It’s widely known (and, I think, pretty uncontroversial) that learning requires effort — specifically, if you don’t have to work at getting the knowledge, it won’t stick.

Even if an LLM could be trusted to give you correct information 100% of the time, it would be an inferior method of learning it.

Exciting news!

Thanks @imvaddi.bsky.social! We're looking for datasets where polymer/molecular structure varies between data points

Do you know of any papers that have datasets of organic materials (polymers, molecules) with experimental parameters–not material properties, but stuff like concentration, temperature, whatever?

#chemsky #matsky

Top row: Daniel Schwalbe-Koda, Shijing Sun, Zakaria Al Balushi, Pieremanuele Canepa, Michael McGuirk. 2nd row: Andrew Zahrt, Cailin Buchanan, Daniel Tabor, James Grinias, Long Luo, Glen O'Neil. 3rd row: Badri Narayanan, Johanna Schwartz, Mark Hendricks, Jessica Sampson, Martin Seifrid.

RCSA, the Arnold and Mabel Beckman Foundation, and the Frederick Gardner Cottrell Foundation have funded seven team projects in the second year of the #Scialog: Automating Chemical Laboratories initiative. bit.ly/4e15oxp

a variable flowcell compared to a fixed lenght cuvette

introduction slide for the project

New preprint online: A flow cell for data validation in computer vision. Computer vision is amazing for reaction monitoring, analytical chemistry, and self-driving laboratories, but it has a flaw… let’s dig deep into this research. #Chemsky 1/n
chemrxiv.org/engage/chemr...

Today's a big day for @ddomlab.org! My first PhD student, Sina Dehghan, presents his research at #ACSSpring2025

Bon résumé.

My kingdom for an edit button

#chemsky #matsky

🚨⏳ Only a few hours left!! Make sure to submit your abstracts to this excellent symposium on "Integrating High-Throughput Techniques and Digital Technologies for #Sustainability" at #ACSSpring2025

bit.ly/3XF7jzU

A publishing platform that places code front and centre Curvenote creates interactive publications based on digital-coding notebooks and aims to increase the transparency and reproducibility of data science. Curvenote creates interactive publications based...

I've had many discussions with colleagues about the future of publishing, especially about how antiquated the current static web page & PDF model is. This write-up in Nature
about curvenote.com is exactly what we need!

www.nature.com/articles/d41...

Surprisingly common in my experience

Agreed! Reposting to annoy my Celsius friends 😉

Gyros PurePep chorus recently installed in the Data-Driven Organic Materials Lab @ NC State

Another instrument is up and running! We set up our
Gyros PurePep Chorus last Friday. Keep an eye out for exciting developments on this front 👀

#chemsky #matsky

We just completed the installation of Big Purchase #1 ™️: a quadruple (!) detector SEC from Tosoh. This bad boy has RI, UV, MALS, & viscometry 💪 We're really excited to start getting some absolute molecular weights!

#chemsky #matsky

Special shoutout to Quyen and her students at UCSB for their dedication to extracting all the data! And to Aspuru-Guzik group students Stanley Lo and Gary Tom for their work putting together a lot of the ML pipeline

We discuss the potential and current limitations of data for OPV devices. We also provide the first dataset of OPV materials and device fabrication parameters, and discuss many of the challenges that face the field due to lax reporting standards.

Beyond molecular structure: critically assessing machine learning for designing organic photovoltaic materials and devices Our study explores the current state of machine learning (ML) as applied to predicting and designing organic photovoltaic (OPV) devices. We outline key considerations for selecting the method of encod...

Our exploration of #ML to predict #OPV device performance from molecular structure of the materials *and* processing data is now officially published in J Mater Chem A as part of their 2024 Emerging Investigators series!

doi.org/10.1039/D4TA...

Heading to my first (!) #S24MRS tomorrow to give an invited talk about machine learning from OPV data and data reporting standards.

Come find me in EN04.02 or drop me a DM if you're in Seattle too!

Across all industries, organizations are rapidly embracing generative AI. Among them, makers of home appliances like fridges and ovens. Generative AI in your oven? Why not? Ater all, AI has been creeping into our homes for years (think smart lightbulbs and Alexa) – but thanks to generative AI, these interactions will become even more human and more personal. Imagine, for example, asking your washing machine whether it’s safe to wash a beloved item of clothing on a certain setting – literally, asking it out loud or via an app. Or you could say to your fridge, “Hey, when am I going to run out of milk?” and it’ll tell you. Integrating generative AI into everyday products could lead to a new era of smart appliances that are not only more adaptive to our needs but also more interactive and engaging.

Hell. You are describing hell. Nobody wants an “interactive and engaging” dishwasher. They just want their dishes cleaned. www.forbes.com/sites/bernar...

IMO, this is the correct response. Trying to figure out if this is chaotic good or lawful neutral...

We provide the first dataset of OPV materials and device fabrication parameters, and discuss many of the challenges that face the field due to lax reporting standards. Special shoutout to the dedicated students who extracted all the data!

Beyond Molecular Structure: Critically Assessing Machine Learning for Designing Organic Photovoltaic Materials and Devices Our study explores the current state of machine learning (ML) as applied to predicting and designing organic photovoltaic (OPV) devices. We outline key considerations for selecting the method of encod...

Out now on ChemRxiv! We explored using machine learning to predict OPV device performance from molecular structure of the materials *and* processing data. Hopefully this somewhat didactic approach is useful to those in the field who are interested in ML.

doi.org/10.26434/che...

#ChemSky #MatSky