Build Night: Web Research Agents That Don’t Break in Prod x HackerSquad
Technology

Build Night: Web Research Agents That Don’t Break in Prod x HackerSquad

SomoEventFinder

Better experience in the app

  • Get a reminder before the event starts
  • Save & share with friends
  • Discover similar events based on your taste

Date

Wed, May 27

Time

12:00 AM - 03:30 AM

Location

625 2nd St

Price

Free

Website

Visit site

Get reminded

Choose when to get reminded by email.

About the event

Join us for an immersive Build Night, where you'll create a robust deep-research agent that outputs structured, citation-backed JSON. This hands-on activity isn't just a talk; it’s your opportunity to build something deployable!

🔍 What We’re Building

  • Project: Deep Research → Structured JSON (with citations)
  • Input: A company product/market question
  • Output: Schema-validated JSON with citations gathered from various public web sources

Once you have your outputs, we’ll switch to scale mode, allowing you to assess how it performs under real-world conditions with concurrency and bounded retries.

⚙️ Challenges to Overcome

If you’ve built agents or RAG before, you know the real challenges lie beyond the model itself—in web handling and operations. Together, we will tackle critical failure modes that often arise in production. Here are some of the issues we'll confront:

  • JS-heavy pages that yield placeholder HTML
  • Inconsistent results across sessions or geographical locations
  • Blocks and friction that only emerge during higher volumes
  • Rate limits causing failure at scale (it works at 1 request per second, but fails at 20)
  • Suboptimal retry behaviors leading to excessive costs
  • Silent extraction failures yielding incorrect structured data
  • Lack of observability, leaving failures and costs unexplained

By the end of the night, you’ll transform your agent from a one-off demo into a repeatable pipeline with measurable success!

🎁 What You’ll Walk Away With

  • A deployable repo (Node/Python starter) featuring:
    • Multi-source retrieval and citation capture
    • Schema-first extraction with validation checks
    • A strategy for browser fallback
    • Session/geo controls for consistency
    • Safe parallelism through concurrency and queue patterns
    • Measurable retry/backoff
    • Basic metrics output capturing success rates and failure reasons
  • Plus, credits to keep your project running after the event.

👥 Who Should Attend

This event is geared toward:

  • Advanced developers working on agents, research workflows, or competitive intelligence
  • AI engineers, full-stack developers, and founders looking to ship notable features
  • Anyone who has ever said, “it worked yesterday” about a web pipeline.

🌟 Don’t miss this chance to elevate your skills and create something impactful!

NOTE: We cannot guarantee the accuracy of the information we provide about this event. Visit the event's website to verify details such as date, opening hours, prices and location.

This week in San Francisco

More events in San Francisco