South Park SF Is Buzzing

I've attended a couple of MLOps events recently at Bright Data's new location around South Park, in SF. Their new offices are primo. I'm struck by the way the sprawling space is used, which says as much about the latest AI tech boom, as anything. I'd say about 1/8th of the space is actually for desks/computers for employees, indicative of how few engineers and employees are needed, on-prem, for the actual product development and support. The remainder is built for event hosting, podcasting (a hollywood-like studio space), and BattleBot matches (they're the official sponsor; I won a match!) I've attended a couple of events there already, in the first month in their new space, they're hosting at least an MLOps event a week, and hackathons over the weekend. They're building a community (and their customers, in the process) as much as a product.

They had a ton of great presenters at the 2 events I attended. Local builders, engineers, VCs, DevRels. Some are obviously in very early stage, some more polished.

But one thing was obvious – so much innovation, so much engineering. There are many new problems to solve with AI. So much opportunity. Some of the problems and showcased solutions included -

1. how do we automate the evaluation of agent output? (evals, evals, evals! + agent work change history) inngest, sutro.sh
2. how do we optimize prompts to get the best performance? (prompt optimizer solutions), sutro.sh, gepa
3. How do we use AI offline? (local models, local rag, so many use cases) offlyn
4. How do we manage massive teams of autonomous agents? (So many agent harness builders right now) agor
5. How do we improve LLM memory? (3 or 4 presenters talking about this one...on everyone's mind right now.) inkeep

Before I attended the first MLOps event, I didn't know anything about Bright Data. I went to get inspired about AI agent building, see what other people are working on, to get some new ideas.

Bright Data is web scraping at scale, as a service. They've positioned themselves as an essential infrastructure/service for AI powered startups and products. If you need data at scale, from the Internet. They might be able to help. I do think their product is pretty cool. And useful. The results of an AI process that uses web tools with Bright Data vs. without, can be stark. I've already started using their service to improve some of of my AI data pipelines, for finance and construction.

I wanted to do something fun with the $50 credit Bright Data gave us for attending the event. I didn't know how this would be received by LinkedIn, since their terms prohibit web scraping, so I didn't post it there (maybe since Microsoft owns it, I'm good? 😉). But I'm going to MSBuild next week. Still a Microsoft guy, for better or worse. Also, still a fan of @steipete, the creator of OpenClaw; he's speaking. I used the Bright Data MCP, and had Claude Code whip up a little tool to connect with and get some insights on people attending MSBuild this year. This list may not be totally comprehensive, I didn't test it rigorously. And if you want your name removed, dm me. But also if you want to show up on the dashboard, drop a post on LinkedIn and use a hashtag like #MSBuild2026, or #MSBuild. It'll pick you up the next day.

goingtomsbuild.netlify.app Who's going to Microsoft Build 2026 84 attendees across 26 countries · pulled from public LinkedIn posts · mapped, broken down by company and geography.

---

Adam Daum is an agentic engineer and AI architect. He runs Weststack, LLC, an agentic AI and software engineering company, and writes about building practical AI solutions at adamdaum.com.