Free Lightning Lessons to Advance Your RAG Implementations¶
I'll be hosting industry experts to share practical techniques for enhancing your Retrieval Augmented Generation (RAG) systems.
Each session will be a live Zoom meeting, offering an opportunity for direct Q&A. You can submit questions during the session, and the most upvoted will be addressed.
What You Will Gain¶
- Insights from Industry Practitioners: Learn from individuals at companies like Cohere, Chroma, Groq, and OpenBB.
- Actionable Content: Each session is designed to provide specific solutions to common RAG challenges.
- Live Q&A: Get answers to your questions in real-time.
- Community Interaction: Engage with other professionals working on similar problems.
- No Cost: All sessions are free to attend.
Session Schedule¶
From Text-RAG to Vision-RAG with VP Search @ Cohere¶
Date: Wednesday, May 14, 2025 at 5:00 PM UTC
Speakers: Myself & Nils Reimers (VP AI Search, Cohere)
Key Learnings: - Core concepts of Visual RAG: Understand how to extend text-based RAG to include visual data. - Architectures for multimodal retrieval: Learn methods for combining text and image understanding in end-to-end systems. - Strategies for production deployment: Discuss practical approaches for implementing and scaling visual search.
Nils Reimers will share insights from his work developing AI retrieval models at Cohere and their application to visual domains. We will examine specific examples of visual RAG implementations and discuss associated challenges and opportunities.
Understanding Embedding Performance through Generative Evals¶
Date: Wednesday, May 21, 2025 at 5:00 PM UTC
Speakers: Myself & Kelly Hong (Researcher at Chroma)
Key Learnings: - Challenges in AI Evaluation: Explore why traditional metrics may not fully capture real-world RAG performance. - Limitations of Benchmarks: Identify potential biases in current embedding evaluation frameworks. - Creating Representative Test Sets: Learn how to develop evaluation datasets that align with production workloads.
Kelly Hong from Chroma will discuss advanced approaches to embedding evaluation. We'll explore how generative evaluations can offer more nuanced insights into embedding quality and cover techniques for implementing robust evaluation pipelines.
Online Evals and Production Monitoring for RAG Systems¶
Date: Wednesday, May 28, 2025 at 5:00 PM UTC
Speakers: Myself & Ben Hylak (Co-Founder, Raindrop)
Key Learnings: - Fundamentals of AI Monitoring: Identify essential metrics for tracking RAG system performance and user satisfaction. - Detecting and Analyzing Issues: Learn techniques for identifying underperformance and diagnosing root causes. - Implementing Feedback Loops: Discuss how to use user feedback for continuous improvement of retrieval quality.
Ben Hylak, Co-Founder of Raindrop, will demonstrate approaches to monitoring RAG systems in production. We will review examples of failures that traditional monitoring might miss and discuss how to implement systems to catch problems proactively.
Latency First: Optimizing RAG & Agent Speed¶
Date: Wednesday, June 4, 2025 at 5:00 PM UTC
Speakers: Myself & Aarush Sah (Head of Evals, Groq)
Key Learnings: - Effective Performance Measurement: Understand tools and methodologies for profiling your RAG stack and identifying bottlenecks. - Identifying Latency Sources: Recognize common performance issues in vector search, context assembly, and LLM inference. - Practical Optimization Techniques: Explore strategies for reducing latency at each stage of the RAG pipeline.
Aarush Sah from Groq will share insights on optimizing inference systems. We will explore techniques with code examples and architectural patterns for reducing end-to-end latency and discuss tradeoffs between speed and quality.
RAG without APIs: Browser-Based Function Calling¶
Date: Wednesday, June 4, 2025 at 6:00 PM UTC
Speakers: Myself & Michael Struwig (Head of AI at OpenBB)
Key Learnings: - Browser-Centric RAG: Learn how to build RAG systems that operate within the browser, potentially reducing backend API needs. - Client-Side Function Execution: Discuss patterns for connecting LLMs directly to browser capabilities. - WYSIWYG Data Integration: Explore techniques for allowing models to interact with on-screen content.
Michael Struwig from OpenBB will demonstrate approaches to building client-side RAG systems. The session will include code examples and patterns for leveraging browser APIs.
Common RAG Anti-patterns and How to Address Them¶
Date: Wednesday, June 11, 2025 at 5:00 PM UTC
Speakers: Myself & Skylar Payne (Founder of Wicked Data LLC)
Key Learnings: - Diagnosing Silent RAG Failures: Identify common failure modes that can impact user experience. - Implementing Robust Monitoring: Learn to build instrumentation to detect RAG system degradation. - Architectural Solutions: Review patterns for addressing common RAG reliability issues.
Skylar Payne of Wicked Data LLC will share experiences from debugging retrieval systems at various scales. You will learn about anti-patterns to avoid and practical solutions.
Natural Language Search on Semi-Structured Data¶
Date: Wednesday, June 18, 2025 at 5:00 PM UTC
Speakers: Myself & Daniel Svonava (CEO at Superlinked)
Key Learnings: - Encoder Stacking: Understand techniques for combining text and metadata in vector search. - Hybrid Data Implementation: Explore approaches to indexing and querying data structures that blend free text with structured attributes. - Advanced Metadata Integration: Move beyond simple filtering to semantic understanding of content and metadata.
Daniel Svonava, CEO at Superlinked, will discuss building ML infrastructure for search that combines structured and unstructured data. He will demonstrate techniques for creating unified embedding spaces.
Improving RAG with Better Data Quality¶
Date: Wednesday, June 18, 2025 at 6:00 PM UTC
Speakers: Myself & Adit Abraham (Co-founder & CEO of Reducto)
Key Learnings: - Precise Data Parsing: Learn techniques for extracting structured information from complex documents. - Optimal Data Representation: Discuss strategies for chunking, indexing, and organizing information for retrieval. - Enterprise-Scale RAG: Explore approaches to handling challenges in large organizations with diverse document types.
Adit Abraham of Reducto will share insights on parsing and extracting data from complex documents. He will demonstrate how data quality impacts RAG system performance and discuss techniques for document processing and knowledge representation.
Session Format¶
Each Lightning Lesson is a 30-minute Zoom session. You can submit questions during the live event, and we will address the most upvoted ones.
This series is part of my work with ImprovingRAG.com to share practical knowledge from leading practitioners.
Registration is required to receive the Zoom link for these free sessions.
You can register for individual sessions using the links above.