From Demo to Production: Deploying AI Responsibly in a Complex World

Dear Data Enthusiasts,
Join us at TU the Sky on the 29th of October to explore how knowledge graphs, robust orchestration, and rigorous fairness testing are shaping trustworthy AI. A big thanks goes to our co-organizers from the AI Factory Austria and Neo4j for sponsoring the event.

Legit Check for AI - Grounding AI with Knowledge Graphs for Transparency and Trust at the Core
AI systems are only as trustworthy as the data they reason over. But when that data is highly interconnected — products, brands, trends, relationships — flat retrieval architectures like RAG fall short: they can't traverse context, can't explain their paths, and can't make their reasoning visible.
This talk shows how Knowledge Graphs change that equation.
Using a real-world dataset scraped from sources like Sneaker News, WWD, and StockX, we demonstrate how to build a Knowledge Graph from unstructured documents and use it to ground Large Language Models, making their outputs not just more accurate, but genuinely explainable.
We'll walk through:

• Building the graph — extracting structured knowledge from unstructured web sources and modeling it as a connected graph
• Grounding LLMs — how Knowledge Graph context replaces hallucination-prone retrieval and gives the model structured facts to reason over
• Visualizing trust — how graph-based visualizations let users see why the AI answered the way it did, not just what it answered

In the sneaker world, a legit check separates what's real from what just looks right. AI deserves the same standard. Along the way, we'll make the case for why Knowledge Graphs outperform RAG and multi-tool agent architectures in domains where data relationships matter and show what trustworthy AI looks like when it can finally show its work.

Moritz Wegener, Data Scientist - ATVANTAGE GmbH
Moritz works on agentic AI, knowledge graphs, and moving AI systems from demo to real-world deployment — including graph-based fraud detection in banking and enterprise-scale decision systems. He'll share what it takes to get agentic AI use cases from prototype to production.

Running Dagster on Slurm: Practical Orchestration for HPC and AI Workloads
Many data and AI teams rely on Slurm-managed HPC clusters, but modern orchestration tools often assume cloud-native infrastructure. This creates a gap: researchers and ML engineers want Dagster-style assets, observability, and reproducibility, while their actual compute runs on shared Slurm clusters.
This talk introduces dagster-slurm, an open-source integration for running Dagster assets on Slurm. We will show how it lets teams develop assets locally, submit production workloads to HPC, manage environments, and surface Slurm job metadata back into Dagster. Using multimodal AI pipelines as a motivating example, we will discuss why orchestration matters when workloads mix CPU preprocessing, GPU inference, embeddings, and downstream analytics. Metaxy appears as one example of a metadata-driven pipeline that benefits from this setup, but the focus is broader: how to make Dagster a practical control plane for Slurm-based compute.
Attendees will learn how dagster-slurm maps Dagster assets to Slurm jobs, how local and cluster execution can stay aligned, and how this approach improves visibility, reproducibility, and resource usage for HPC-backed data workflows.

Georg Heiler, Senior data scientist
Georg is a co-founder @Jubust and a Senior data expert at Magenta as well as a ML-ops engineer at ASCII. He is solving challenges with data. His interests include geospatial graphs and time series. Georg transitions the data platform of Magenta to the cloud and is handling large scale multi-modal ML-ops challenges at ASCII.

Testing AI for Fairness in Practice
Drawing on real-world cases, from biased healthcare algorithms to discriminatory tax audits and predictive policing, this talk unpacks what "bias" actually means in AI systems and why fixing it is harder than it looks. Rania Wazir, Co-founder & CTO of leiwand.ai, walks through the practical mechanics of bias testing: identifying at-risk groups, choosing the right fairness metrics, and confronting the uncomfortable trade-offs between accuracy and fairness.

Rania Wazir, CEO at Leiwand
Rania Wazir is co-founder and CTO of leiwand.ai, a Vienna-based startup helping companies and organizations develop and deploy trustworthy AI. A mathematician by training, with a B.Sc. from Stanford and a Ph.D. from Brown University, Rania has spent years as a data scientist specializing in trustworthy AI, NLP, and bias detection in algorithmic systems. She led a consortium investigating algorithmic bias for the EU Fundamental Rights Agency and serves as tech lead on an Austrian Research Agency-funded project building a "fair by design" AI development process. leiwand.ai's work includes supporting the EU Commission's rollout of the AI Act across Europe. Rania is also active in international AI standards bodies (ISO/IEC, CEN/CENELEC), bringing a rare combination of technical depth and regulatory fluency to questions of AI accountability and the EU AI Act.

🎤🎤 Open Mic
We are going to open up the stage after the talks for community announcements. If you'd like to announce something, open this slide deck, make sure you are signed in with a google account, and click "View Only" -> "Request Edit Access". Explain in the text box what you want to announce, and we'll give you edit access to the slide deck.
🎤🎤
We’ll have some food and drinks after the event.
Please note that during the event, photos might be made and later posted on VDSG's social media page. Please notify us if you do not agree.
Attention attendees with food allergies. Please be aware that the food and drinks provided may contain or come into contact with common allergens, such as dairy, eggs, wheat, soybeans, tree nuts, peanuts, fish, shellfish, or wheat.

Best,
The Organizer Team