​The era of AI demos is over. The Munich AI Nexus is a senior-level meetup series for architects and engineers solving real-world production AI challenges.

​At the Kickoff: During the 'From Hack to Scale' session, we asked: 'What’s actually breaking AI in production?'
The results were clear:

• ​42% said 'context' is the most overlooked architectural layer,

• ​41% flagged 'data quality and chaos' as their biggest blocker.

​MEETUP FOCUS:

• ​Theme: "Context – From Data Chaos to Clarity"
  This module focuses on transforming raw, fragmented data into structured, trustworthy context for AI systems.

• ​You can't automate what you don't understand. Agents without clarity are just more expensive hallucinations.

• ​Upvote the topic-related issues you find the most interesting.

• ​Read on here to discover the 9 themes for our upcoming meetups and the final scope of our Production Readiness Blueprint.

​SUBMIT YOUR CASE:

​Share your toughest "Data & Context" challenge (even rough!)

​Examples:

• ​Keeping customer context consistent across dozens of systems (CRM, ERP, ticketing)

• ​Enforcing GDPR/PII compliance in multi-tenant RAG without killing personalization

• ​Debugging and auditing contextual reasoning so compliance teams trust outputs

​AGENDA

​18:00 - 18:05 - Welcome

​18:05 - 18:15 Recap & Opening

• ​Survey recap: 42 % flagged context as the most overlooked layer; 41 % cited data chaos as the top blocker.

• ​Focus for tonight: moving from hacky demos to scalable, trustworthy AI.

​18:15 – 18:45 - Talk 1: Daniel Töws, codecentric AG - LLMs vs. Messy Data – Lessons from the Trenches

• ​Chaotic CSVs → structured data: 30 K attributes, iterative refinement, and Pydantic.

• ​Agentic Data Buddy: navigating 7 K tables to answer “Why aren’t these pants selling?” — balancing speed vs. trust.

• ​Open question: Can we trust LLM-generated analysis at scale?

​18:45 – 19:15 - Talk 2: Alexander K, Horsch - plot-RAG (pRAG): Visualizing & Optimizing RAG Performance – Smarter RAG Evaluation

• ​Visualizing retrieval bottlenecks to optimize speed and accuracy.

• ​Synthetic datasets that rival manual evaluation — GPTMB 2025 Best Paper Award.

• ​Formula: defining the optimal retrieval window for performance vs. precision.

​19:15 – 19:45 - Discussion: Debugging Context in Production

• ​Case study: data silos, compliance, and trust.

• ​Stakeholder POV: cultural friction — “We’ve always done it manually.”

• ​Audience: upvote and discuss top “Data & Context” challenges.

​19:45 – 20:00 - Closing & Next Steps & Networking

• ​Submit your toughest challenge for the Production Readiness Blueprint.

• ​Preview next meetup: Behavior – From Heuristics to Planning.

• ​Host: Asaad Almutareb [founder of artiquare, appliedAI Developers MUC Meetup Curator]

• ​The appliedAI Developers bring together engineers, researchers and tech leaders solving real-world AI challenges through knowledge exchange and collaboration.

• ​Codecentric AG develops software solutions for the future, something everyone of our 500+ employees in each of our 13 locations in Germany is passionate about.