LLM FinOps

Vibe coding gives you momentum; unmonitored tokens give you a runaway invoice. When 85% of teams overshoot their AI budgets, cost isn't a finance problem—it’s an architectural defect driven by bloated context, redundant tool calls, and infinite agent loops.
In this session, we break down how to treat LLM FinOps as an active engineering discipline:

• Context Engineering over Token Bloat: Decoupling knowledge into a deterministic Context Catalog to eliminate multi-hop overhead and slash query costs by up to 80%.
• Semantic Caching & Smart Routing: Stopping repeated prompt fees and dynamically offloading routine tasks from frontier models to SLMs.
• Governance & Guardrails: Using AI Gateways and local circuit breakers to catch non-deterministic loops before they drain credits.
• Unit Economics & Observability: Tracking cost-per-capability and token health directly in your engineering workflow.

Join us to learn how to sustain development velocity while keeping inference costs predictable and production-ready.

⏲️ Agenda
18.30* Welcoming
18.45 Talk
19:30 Q&A
19:45 Closing remarks & Networking and see you at the next Meetup!

*You will receive detailed information on how to access the building close to the date.

🎙️Speaker
Alessio Bertazzo - Senior Principal Engineer @ Mia-Platform
A Software Engineer with solid roots in Java and Spring/Spring Boot backend development, he currently holds the position of Senior Principal Engineer in Mia-Platform's R&D team. In this role, he helps define the technical and strategic vision of the products, bridges the gap between the Product division and other company departments, and drives innovation with a strong focus on Generative AI.