Memory-Aware Agent Harnesses for Anti-Money Laundering — Nacho Martinez, Oracle

By ai.engineer

Categories: AI, Tools

Summary

Building AI agents requires focusing on the harness layer—memory, context engineering, and data manipulation—rather than frontier models alone. Nacho Martinez shows how autonomous agents solving complex problems need custom memory systems and human-in-the-loop mechanisms that LLMs don't inherently possess, demonstrated through an anti-money laundering use case.

Key Takeaways

  1. Agent autonomy is limited by context windows and missing memory components. Builders must implement custom memory systems to enable agents to work autonomously for 10-100+ hours on complex problems.
  2. The agent stack has five layers: application interface, model reasoning core, infrastructure, compute, and data. The data layer is most customizable and controllable—especially critical when using third-party models.
  3. Focus engineering effort on agent harnesses, not just frontier models. The harness (memory, context engineering, data manipulation) is where product differentiation and customization actually happens.
  4. Real-world agent applications require human-in-the-loop mechanisms. A bank's anti-money laundering system should flag suspicious patterns (like $1M deposit followed by $50K ATM withdrawals) to reduce human review time.
  5. Agent applications solve practical compliance problems at scale. Medium-sized banks can autonomously categorize suspicious transactions into five types, freeing compliance officers from manual review.

Related topics

Transcript Excerpt

[music] Hi. Hi everyone. Uh thank you for coming to this session. Uh we're it's going to be a session about agent harnesses. We don't have a lot of time. It's only one hour session. So um hopefully we have time to talk about a lot of things that that you guys are interested in and the the session essentially is going to be uh like me talking like 10 15 minutes about the topics that we're going to talk about and solve in the workshop. You have the workshop instructions in a QR code or in a website and if you don't want to listen to me, you think I'm boring or something, you can start uh on your own on the workshop as I'm speaking. But essentially this is this is what's uh happening on the website. You just um sign up with your GitHub handle and then you will automatically receive an invitat…

More from ai.engineer