Disclaimer: This summary has been generated by AI. It is experimental, and feedback is welcomed. Please reach out to info@qconlondon.com with any comments or concerns.
Presentation Title: Achieving Precision in AI: Retrieving the Right Data Using AI Agents
Speaker: Adi Polak
Overview: This presentation focuses on improving the precision of AI systems, especially in the context of Retrieval-Augmented Generation (RAG) and the development of agentic systems to manage data retrieval and processing with greater accuracy.
Key Points:
- Challenges of Precision: Emphasizes the difficulty in achieving precision in AI systems, which is crucial for transitioning from prototypes to production-grade AI systems.
- Agentic RAG: Introduces agentic RAG as an advanced AI architecture that improves precision by segmenting data storage and retrieval for training vs. inference. This approach reduces latency and costs.
- Component Needs: Highlights the need for applications that interact with LLMs while utilizing short-term and long-term memory, caching capabilities, and intelligent feedback loops.
- Integration and Optimization: Discusses the integration of multiple data retrieval methods, including term search and similarity search, to enhance data accuracy and retrieval speed.
- Agent Functionality: Describes the functionalities of AI agents, including planning, acting, and use of feedback loops to improve decision-making processes.
- Precision Metrics: Emphasizes the importance of specific metrics for evaluating the precision of AI models, including metrics for hallucination rates and continuous feedback loops.
Conclusion: The presentation underscores the necessity of precise data retrieval for the efficacy of AI systems and highlights agentic RAG as a pioneering approach to enhance accuracy while minimizing resource use and operational costs.
This is the end of the AI-generated content.
In the race to harness the power of generative AI, organizations are discovering a hidden challenge: precision. Models are only as effective as the data they access, yet most approaches to Retrieval-Augmented Generation (RAG) lack the dedicated, fine-tuned pipelines needed to ensure the right information is delivered at the right time.
Today, most RAG systems pull from vast, generalized data lakes, leading to noisy outputs and frustrating inefficiencies. The result? Wasted resources, inconsistent responses, and missed opportunities for real-time decision-making. But what if you could create an AI system that doesn’t just retrieve data—but understands its context, delivering precise, actionable insights in milliseconds?
This is where agenticRAG comes into play—a breakthrough in AI architecture that pairs dedicated retrieval pipelines with intelligent agents to deliver pinpoint accuracy. By segmenting your data storage and retrieval processes specifically for training vs. inference, you can achieve hyper-focused precision while dramatically reducing latency and costs.
Imagine an AI system that knows exactly what data it needs and how to get it with zero lag—a system that’s tuned to perform like a well-trained expert in your domain.
Curious to discover how you can optimize your AI applications for laser-focused accuracy? Join me as to learn more about AgenticRAG and fine tuning your models.
Speaker

Adi Polak
Director, Advocacy and Developer Experience Engineering @Confluent, Author of "Scaling Machine Learning with Spark" and "High Performance Spark 2nd Edition"
Adi is an experienced Software Engineer and people manager. She has worked with data and machine learning for operations and analytics for over a decade. As a data practitioner, she developed algorithms to solve real-world problems using machine learning techniques while leveraging expertise in distributed large-scale systems to build machine learning and data streaming pipelines. As a manager, Adi builds high-performance teams focused on trust, excellence, and ownership.
Adi has taught thousands of students how to scale machine learning systems and is the author of the successful book Scaling Machine Learning with Spark and High Performance Spark 2nd Edition.