Dude, where did my tokens go?
MCLD 3002 | Sat 08 Aug 2 p.m.–2:45 p.m.
Presented by
-
Nikola Grcevski has worked as a software engineer for more than 20 years, mostly in the field of compilers, managed runtimes and performance optimization. Most recently he's working on low level application instrumentation with eBPF at Grafana Labs. He's currently a maintainer of two OpenTelemetry projects: OpenTelemetry eBPF Instrumentation and OpenTelemetry Injector.
-
Rafael is a software engineer who has been building systems software in C++ and contributing to open source projects since 2005. He is a maintainer of OpenTelemetry eBPF Instrumentation (OBI), where he works on eBPF-based application observability and auto-instrumentation. Prior to OBI, he contributed to Grafana Beyla before its donation to the OpenTelemetry project.
Earlier in his career, Rafael worked extensively with the Qt project, serving as the maintainer of Qt's QNX integration plugin and contributing to QtCore, QtMultimedia, Qt Creator, and qmake. While at Blackmagic Design, he architected a declarative UI framework for the company's embedded real-time operating systems, laying the foundation for interfaces used across products including digital cinema cameras and ATEM switchers.
His work has often sat at the boundary between applications and operating systems, spanning embedded user interfaces, developer tools, kernel-level instrumentation, and observability systems that help explain what software is really doing.
Nikola Grcevski has worked as a software engineer for more than 20 years, mostly in the field of compilers, managed runtimes and performance optimization. Most recently he's working on low level application instrumentation with eBPF at Grafana Labs. He's currently a maintainer of two OpenTelemetry projects: OpenTelemetry eBPF Instrumentation and OpenTelemetry Injector.
Rafael is a software engineer who has been building systems software in C++ and contributing to open source projects since 2005. He is a maintainer of OpenTelemetry eBPF Instrumentation (OBI), where he works on eBPF-based application observability and auto-instrumentation. Prior to OBI, he contributed to Grafana Beyla before its donation to the OpenTelemetry project.
Earlier in his career, Rafael worked extensively with the Qt project, serving as the maintainer of Qt's QNX integration plugin and contributing to QtCore, QtMultimedia, Qt Creator, and qmake. While at Blackmagic Design, he architected a declarative UI framework for the company's embedded real-time operating systems, laying the foundation for interfaces used across products including digital cinema cameras and ATEM switchers.
His work has often sat at the boundary between applications and operating systems, spanning embedded user interfaces, developer tools, kernel-level instrumentation, and observability systems that help explain what software is really doing.
Abstract
There’s nothing easy about GenAI Observability. GenAI observability is a moving target, new frameworks and SDKs appear daily, while organizations are rapidly integrating AI into their products. So how does instrumentation keep up with this ever expanding ecosystem of tools and libraries?
While there are many new frameworks, libraries and SDKs to build GenAI capabilities in applications, there are only a handful of GenAI vendors that all these libraries use. In this talk, we show how recent updates to OpenTelemetry eBPF Instrumentation enable zero-effort, cross-language GenAI observability by capturing signals at the protocol level, regardless of the libraries or frameworks in use.
We’ll demo how to capture GenAI signals for core AI agent and MCP server operation types, e.g. tool call, embedding, rerank, agent invocation, prompt management and resource access, across multiple GenAI vendors all with OpenTelemetry eBPF Instrumentation and 100% programming language agnostic.
There’s nothing easy about GenAI Observability. GenAI observability is a moving target, new frameworks and SDKs appear daily, while organizations are rapidly integrating AI into their products. So how does instrumentation keep up with this ever expanding ecosystem of tools and libraries?
While there are many new frameworks, libraries and SDKs to build GenAI capabilities in applications, there are only a handful of GenAI vendors that all these libraries use. In this talk, we show how recent updates to OpenTelemetry eBPF Instrumentation enable zero-effort, cross-language GenAI observability by capturing signals at the protocol level, regardless of the libraries or frameworks in use.
We’ll demo how to capture GenAI signals for core AI agent and MCP server operation types, e.g. tool call, embedding, rerank, agent invocation, prompt management and resource access, across multiple GenAI vendors all with OpenTelemetry eBPF Instrumentation and 100% programming language agnostic.