Jongmin (Jeongmin) Baek

speaker

142 appearances 1 recordings 1 series first heard Jun 2026 last heard 30 Jun

Jongmin (Jeongmin) Baek’s voice in public audio — every appearance, attributed to the second.

Trend

recordings per month · last 12 months

Recordings per month over the last 12 months — 1 in all, peaking in Jun 2026 with 1.

Appearances

newest first · ▶ plays the moment

Working Smarter · How agentic AI works behind the scenes to find the answers you need · 30 Jun 2026 podcast

A lot of new paradigms come and go.

And for a team that tries to build a meaningful AI product on top of these innovations, you have to carefully balance...

building good foundations, and also making sure that you adopt the new paradigms so that you can shape silly dairy things.

Typically, these innovations happen over years and decades, right?

And there is time for both academic labs and industry to do benchmarking, you know, go down a path and decide that this is not it, or maybe this is it.

But all that is compressed into spans of months.

So that's why I call it amorphous to some extent, because there is no playbook.

So it's been up to Jongmin and his team to help write it.

We are trying to build an agent that tries to extract insights from your ongoing projects and your collaborators.

We have agents that tries to summarize what has happened since the last time you logged on.

And there are other agents that, you know, interact with Dropbox as a file system.

You could tell it to say, you know, every time PDF is uploaded, you know, summarize it and then put the total from a receipt in this Excel sheet and things like that.

One of the interesting complexities of building and shipping agents, which is that none of this is deterministic.

Yes.

You're talking to an LLM and giving instructions through natural language for the most part.

So this means that you need to have evaluation guardrails that basically tells you how often and how well an agent follows the instructions that you expect it to follow.

and how good the outputs or the outcomes are of what the agent has returned as a message or the actions that's taken.

There are several paradigms for evaluating agents and their executions.

So one is the most expensive, but it's the gold standard, which is humans, right?

So you ask people to evaluate the particular thing that AI did and say, is this good?

Showing 21–40 of 142 · page 2 of 8