Abstract
We’re all excited to build and deliver agentic AI services. But what about running at the exponentially greater scale that agents create? LLMs suffer from poor latency and availability issues. More frequent model training drives more frequent updates to agentic services. Most of all, the LLM cost of running at an agentic scale breaks the bank—fast. So, what can you do?
In this session, we’ll dig into how engineering and operations can address:
- Making agentic services fail-proof when their LLMs are not
- Managing a two-order-of-magnitude increase in TPS, including a 2M TPS RAG case
study - Navigating cost vs. quality tradeoffs, with LLMs costing up to 100,000x more than a
database transaction - Continuously redeploying agents that require frequent retraining
After this session, we invite you to attend part 2 of the discussion: “From Concept to Code: Navigating Agentic AI Services"
Sponsored session
Session sponsored by
Akka is used to develop resilient, low latency, large scale, cloud-to-edge distributed applications.
QCon London 2025 is a three day conference for senior software engineers, architects and team leads. An international program committee of working engineers selects every session. Patterns and practices, not products and pitches.
Part of the track
Sponsored Solution Track IFrom the same track
Monday 7 April
11:45 Westminster (4th Fl.) Sponsored Zero Trust or Bust. Bring Your Own Cloud To Real-Time Data Karin Landers Senior Product Marketing Manager @Ververica As organizations increasingly look to adopt Bring Your Own Cloud (BYOC) models to deploy scalable, real-time data streaming solutions, security concerns around data governance, access control, and compliance become critical business considerations. 13:35 Windsor (5th Fl.) Sponsored A Blueprint for Agentic AI Services Duncan DeVore Sr. Director & Architect Advocate @Akka We’re all excited to build and deliver agentic AI services. But what about running at the exponentially greater scale that agents create? LLMs suffer from poor latency and availability issues. More frequent model training drives more frequent updates to agentic services. 14:45 Westminster (4th Fl.) Sponsored Building a Streaming Agentic AI Pipeline with Redpanda and Snowflake Paul Wilkinson Principal Solutions Architect @Redpanda In this technical talk for developers, architects and the technically curious, Paul will cover recent developments within Redpanda Connect. 17:05 Westminster (4th Fl.) Sponsored AI Developer Tools Are Focused on the Wrong Problem Dennis Pilarinos Founder & CEO @Unblocked For all the claims about AI increasing developer productivity, why aren't developers seeing more of an impact?What we've heard from thousands of developers is that writing code is not their main challenge.