Hey!

We're Rasyn, and we built a platform called Marigold. Give it any chemistry task in plain language and it pulls the literature, reads your files, picks the tools, sets up and runs the calculations on its own machine, and hands back the numbers along with its reasoning what it did to get them. 

It's live at app.rasyn.ai; here's a short demo: https://youtu.be/ru1BeRBlRec

Here's the team:
Our team has worked at NASA JPL, and researched with scientists from Nvidia and Deepmind. Our team has worked on softwares serving 50 million users, and their research was used by frontier AI labs including OpenAI and Meta. Our team studies Computational Chemistry at Caltech, has done Total Synthesis for 4+ years, and researched alongside Nobel Laureates, so they have faced these problems firsthand. They'd lose hours deciding which tool to call next, moving outputs between disconnected tools, and running a single program at a time.

Problem:
Chemistry has countless software. There's one for predicting a structure, another for proposing a route, another for reading a spectrum. None of them talk to each other, so you're stuck feeding outputs between them; this clerical work is where the time goes, none of which is science.

**Solution:**
We tried coding agents first. They're built to write software, so a chemistry task gets handled as a scripting problem and the chemistry goes unverified. Marigold checks between steps and stops when something doesn't hold.

It runs on its own machine, so it installs and executes what it needs. It works with the tools chemists already use, submits jobs to your cluster, and runs on whatever model endpoint you have, so your data can stay in the building. In the demo we hand it four thousand compounds and ask which are worth screening. It does the whole thing itself.

Longer term we're building toward a general intelligence for chemistry. Marigold today reasons over existing tools, and that's where we start.

The Ask:

This is only the first version; it will break. And we'd like it to break in front of chemists, so if you run any sort of chemistry, please try it and tell us where it falls over. 

We'd also love intros to any lab doing chemistry, computational or wet, since we want to sit down and work through a real problem with people. 

Happy to answer any questions!