TDWI Konferenz 2024 - Insights and Reflections
Reviewing TDWI Konferenz from 11.-13. June 2024 in Munich, Germany
Again I attended the TDWI Konferenz in Munich this year. It is often called “The class reunion for data & analytics in Germany“. My employer INFOMOTION had a both, a Special Day and we delivered some presentations like
RAG Architecture and LLMOps by my colleagues An Dang and and Asame Suud
Value Exploration for Data Products by my colleagues Alexander Bauer and Laura Weber
Data Security - Impact of Gen AI on Data & Analytics by my colleage Johannes Wenzel
Data Fabric by myself :-)
Fig. 1: Impressions from TDWI conference 2024
This means, my company and myself as a TDWI member are very engaged for TDWI and after getting access to the presentations I want to reflect what was hot, what not and what brought new insights.
I asked some currently popular AI’s what they identify as the top 5 topics from the conference, based on around 160 presentation titels I found:
Fig. 2: What AI’s identify as main topics of the conference
but also this Word Cloud shows, what it was about - Data, AI, Analytics as big topics, GenAI & LLMs, Data Architecture, Data Warehouse, Data Mesh, Data Vault, Data Governance, Data Catalog, Data Culture, ESG and so on:
Fig. 3: Word Cloud based on the titles, (slightly harmonized)
So far I collected some insights and take-aways from the 3rd day of TDWI. Therefore later there will be a part 2 für the first two days or I expand later here:
Strategies for a Seamless Data Shoping Experience (Camelot)
“We strive to become a data-driven organization, where shopping for data products is as easy as picking groceries.”
Data Assets become Data Products through Data Governance
Data Products foster collaboration
Data Consumption becomes a user-friendly experience thanks to the Data Marketplace
Open Source Data Platform in Healthcare (infologistix)
Advantages of OSS: Cost-effective scaling, cost-effective flexibility, speed of innovation
Disadvantages of OSS: Complexity & compatibility, enterprise support, "perceived security concerns"
Corporate Analytics as a Navigation Tool (Hapag-Lloyd)
Understanding Data & Analytics in the company as a ongoing communication and active change management using:
Reoccurring intranet articles
Bi-weekly open-office sessions
Community & developer live streams
Events & internal presentations
Expert-driven community
Finally be Heard and Understood with Data Storytelling (Data Story Lab)
If decision-makers don't understand your data project, they don't need more information. They need less of it.
You want more impact? Create images in people's minds instead of on dashboards.
When you talk about your data project, don't talk about data, talk about people.
Technical Debts (BARC)
Causes: Deadline and cost pressure, M&A, quick fixes, shadow IT, workarounds, changed roadmaps of manufacturers, knowledge gaps, missing design patterns, ...
Consequences: High susceptibility to errors, long development times, additional costs, loss of trust, ...
Approach: Record TD, evaluate, plan rectification and proactive handling of TD
GenAI - A Field Report (iteratec + BAM)
Using GenAI as part of the process for concrete formulation process für sustainability:
Context Driven instead of Data Driven / Knowledge Mining / Trigger Words
Using Zero Shot Design - but not useful to consider new insights
Using a Verifier Model for forward prediction of compression strength to integrate feedback
Insight: LLMs can automate gut feeling
Privacy-Aware Retrival Augmented Generation (CROZ)
Use Lexical Search for: very fast, exact matching, short queries, interpretability
Use Semantical Search for: semantic context understanding, multi-lingual, multi-modal, longer queries
Considering to combine both - hybrid search on RAG for security trimming
Consider rules for Responsible AI:
Inform user that they are talking with AI system
Inform user about possible hallucinations or other problems
with LLMs
Inform user about how data is being used
For every generated answer show related documents used for
answer generation