What should I attend at MLOps World? We had bem read every talk

There are very few things I enjoy more than attending conferences.

And the main reason is this: conferences are treated by us as user research, not as marketing. So much valuable feedback is gotten from the market in a couple of days on a show floor. You hear what's broken in people's stacks, what they've tried, and what they've given up on. You can't get that from a dashboard.

That's why the core team goes to every conference. We don't hire salespeople or hired guns to staff the booth. It's always the founders plus more people from the core team, the same people who build and run the product.

It's also the reason there's no looping video at our booth. We have the real product. Every demo you get from us is a real demo of the real product.

Austin, next week

We are attending MLOps World in Austin next week, which we're very excited about. Austin is a city I've never been to. It's actually going to be my first time in Texas at all, and I'm already incredibly excited to try the barbecue.

So we were thinking as a team: what could be a very cool way to show how bem works, for a conference we're about to attend?

And what we thought is: why don't we create a little tool that lets people take what they know, what they're learning and what they care about, and match that with all the incredible talks and workshops that are going to be at the conference.

So that's what we built.

Find your talks

Pick the topics you care about and what you do, or type anything (a tool, a model, a problem you're stuck on). The sessions get ranked, and every result tells you why it matched. Open any of them to see exactly what bem extracted from it.

Or paste your LinkedIn URL. Your profile is read by bem, your skills are pulled out, and they're matched against the skills and topics of every session. LinkedIn doesn't always let us read a profile, so if we get blocked, the tool asks for the PDF instead (on your profile: Resources, then Save to PDF). Your profile isn't saved by this page. Only your roles, topics and skills come back.

How it was built

It took one workflow and 29 seconds.

• Every talk and workshop was pulled from the MLOps World site on November 9: 65 sessions over four days, from 68 speakers.
• Each session was sent through a bem workflow with one extract function. Out came the topics (from a fixed list of 15), the concrete skills and tools it covers, the level, who it's for, and one line on what you walk out with.
• All 65 were run in parallel. The whole agenda was indexed in 29 seconds.
• The LinkedIn side is a second workflow: one more extract function that reads a profile (the PDF or the public page) into roles, topics and skills. There's no field for your name or contact details in that schema, so they're never extracted.

The matching itself is plain scoring in your browser. Topic overlap counts the most, then skills, then your role, then anything you typed. No model is used in that last step, so you can see why every session is where it is.

This is the same thing teams use bem for every day, just pointed at a conference agenda instead of invoices or claims. Teams use bem to build AI-native products that understand any kind of data, make multi-layered decisions, and leave an audit trail on every run. Here the data is 65 abstracts, the decision is "should I be in that room," and the audit trail is the "see what bem extracted" link on every result.

Where to find us

• Monday, November 16, 11:30 AM to 12:00 PM CT, virtual day. I'm giving a talk: "There's No One There to Press Retry: What We Learned Running AI on a Million Documents a Day." It's the honest version of the climb from 80% to the 99%+ that operational systems need, including two of our own failures.
• Monday, November 16, from 6 PM. Our happy hour with Runpod.
• Tuesday, November 17 and Wednesday, November 18, 8:30 AM to 4 PM. Our booth, table 5 at the Etter-Harbin Alumni Center. It's the first table after registration. Come get a real demo.
• Thursday, November 19, 11:15 AM to 12:45 PM. Upal and Matt run our workshop: "Make It Never Fail: A Hands-On Lab in Taking AI Extraction from 80% to Production-Grade." A working pipeline is stood up in the first fifteen minutes, and then an hour is spent breaking it the way production does.

If you're going, use the tool, build your schedule, and come say hi. And if you know where the best barbecue in Austin is, please tell me at the booth.