The Hidden Layers of AI Systems: Memory, Retrieval, and Search
About Event
Most AI systems don’t fail because of models — they fail because of memory, retrieval, and evaluation.
Agents accumulate too much context and lose signal. RAG pipelines degrade due to poor chunking. Search systems return results, but not always the right ones. And evaluating retrieval quality remains one of the hardest problems in production AI.
This meetup explores what it actually takes to build reliable, production-ready AI systems, with a focus on:
how agents store, retrieve, and forget information
how to improve context quality in RAG systems
how to evaluate search performance beyond simple metrics
how embeddings behave outside traditional text-based use cases
We’ll hear from engineers working on real-world AI infrastructure, followed by Q&A and time to connect with other builders in Amsterdam.
Topics We’ll Explore
Agentic Memory Architecture
Why memory matters in production AI systems
Avoiding context bloating and managing long-term memory
When agents should forget: memory lifecycle, drift, and degradation
Practical patterns using persistent memory systems (e.g. mem0 with Qdrant)
Late Chunking in Production RAG Systems
Why naive chunking leads to poor retrieval
Moving from sentence-level embeddings to contextual chunking
Preserving document-level meaning across chunks
Improving recall and relevance in real-world systems
Evaluating Search Quality
Why evaluation is often overlooked in retrieval systems
Comparing exact match vs semantic (fuzzy) retrieval
Practical mental models for measuring search performance
Insights from recent research on retrieval evaluation
Embeddings Beyond Text
Working with embeddings for code, tables, time-series, spatial, and image data
Retrieval patterns for non-text modalities
Real-world example: computer vision at the edge
Multi-camera product verification and semantic search across visual datasets
Agenda
5:30 – 6:15 PM | Doors Open — Arrival & Networking
Check-in, food, drinks, and informal networking
6:15 – 6:25 PM | Welcome & Kickoff
Opening remarks from MLOps Community
6:25 – 7:05 PM | Qdrant
[Talk Title TBC]
Deep dive into memory architecture, retrieval quality, and evaluation in production AI systems
7:05 – 7:35 PM | Speaker TBC
[Talk Title TBC]
7:35 – 8:05 PM | Speaker TBC
[Talk Title TBC]
8:05 – 9:00 PM | Panel Discussion, Audience Q&A + Networking
Shared discussion across speakers, audience questions, drinks, and networking
Who Should Attend
AI / ML engineers
Backend and data engineers
Engineers building RAG or agent systems
Technical founders and product builders
Anyone working on search, retrieval, or AI infrastructure
Seats are limited.