Uploaded February 2026 | Updated September 2026, 2 weeks ago
by Maarten Vandeperre
That bill from the Gringotts of third-party LLMs just arrived via owl post. It's... a howler. And the Aurors from your security department are asking pointed questions about which dark wizards are seeing your company's secret spells. In the rush to use powerful magic, we've dabbled in the Dark Arts of black-box APIs, trading control for convenience and inviting Dementors of risk into our infrastructure. It's time to master our own magic.
This session is for wizards of engineering who are ready to move beyond chanting borrowed incantations. We'll learn the potions and charms needed to architect a secure, private model-as-a-service (MaaS) platform using powerful, open-source magic. This isn't Transfiguration theory—it's a practical guide to creating your own Marauder's Map for models. Using API Connectivity as our all-seeing eye, we'll cast protective enchantments (access policies), consult the Pensieve of analytics, and manage our galleon spend.
Leave this session with your own spellbook to stop being a squib (i.e., a person who was born into a wizarding family but does not possess any magical powers) and start being a Master of the AI Arts. It's time to take back control, brew your own powerful potions, and deploy your models with the confidence of Dumbledore himself.
by Maarten Vandeperre
That bill from the Gringotts of third-party LLMs just arrived via owl post. It's... a howler. And the Aurors from your security department are asking pointed questions about which dark wizards are seeing your company's secret spells. In the rush to use powerful magic, we've dabbled in the Dark Arts of black-box APIs, trading control for convenience and inviting Dementors of risk into our infrastructure. It's time to master our own magic.
This session is for wizards of engineering who are ready to move beyond chanting borrowed incantations. We'll learn the potions and charms needed to architect a secure, private model-as-a-service (MaaS) platform using powerful, open-source magic. This isn't Transfiguration theory—it's a practical guide to creating your own Marauder's Map for models. Using API Connectivity as our all-seeing eye, we'll cast protective enchantments (access policies), consult the Pensieve of analytics, and manage our galleon spend.
Leave this session with your own spellbook to stop being a squib (i.e., a person who was born into a wizarding family but does not possess any magical powers) and start being a Master of the AI Arts. It's time to take back control, brew your own powerful potions, and deploy your models with the confidence of Dumbledore himself.
![[VDBUH2026] Magda Miu & Alin Miu - Engineering Leadership in the Age of AI](https://i.ytimg.com/vi/DAgVdXO4NFM/mqdefault.jpg "[VDBUH2026] Magda Miu & Alin Miu - Engineering Leadership in the Age of AI
Engineering management isn’t a promotion, it’s a career shift from solving technical problems to navigating human complexity. In the AI era, this shift becomes even more challenging as leaders must make high-stakes decisions under uncertainty and guide teams through rapid change.
This talk is a condensed “management lab” for senior engineers and new managers, offering practical frameworks for building psychological safety, running effective feedback systems, coaching for autonomy, and making AI-aware decisions.
You will leave with actionable tools, not theory, to help you lead with clarity, confidence, and strategic impact from day one.")
![[VDBUH2026] Abdel Sghiouar - Optimizing LLM Inference for the Rest of Us](https://i.ytimg.com/vi/G58PbxBXC8c/mqdefault.jpg "[VDBUH2026] Abdel Sghiouar - Optimizing LLM Inference for the Rest of Us
Not every organization operates with the hyperscale resources of Anthropic, Google, or OpenAI. For the majority of businesses integrating Large Language Models (LLMs) into their critical paths, the high costs and scarcity of GPU/TPU accelerators present a significant challenge. Striking the balance between performance, availability, scalability, and cost-efficiency is a must.
While Kubernetes is a ubiquitous runtime for modern workloads, deploying LLM inference effectively demands a specialized approach. This session dives deep into practical strategies for optimizing your Kubernetes clusters and LLM Inference workloads to run efficiently and cost effectively. We will explore:
– Container and Model Optimization
– Accelerator Management
– Data & Storage
– Network & Load Balancing
– Observability
Attendees will leave with practical techniques for maximizing cost/performance for LLM inference for their AI-powered applications on Kubernetes.")
