Following on the previous talk (Four Pillars of AI at Enterprise Scale) Petar Zrinscak shows what the four pillars look like once a real organisation tries to stand them up. Drawing on CROZ’s experience building the Jarvis AI platform — tuning vLLM on H200s with Mixture-of-Experts models, designing for governance and observability from day one, and the open-source benchmarking framework the team built along the way — Petar walks through the practical side: what worked, what didn’t, and what they’d do differently. We close with a joint Q&A.
If you’re responsible for a platform that will host AI workloads — and especially if you’re not an AI specialist — you’ll leave knowing how the pillars fit together and exactly what your AI team means when they start throwing acronyms around.
