Opening Pandora’s Box: Building Effective Multimodal Feedback Loops
🎥 Recorded live at the MLOps World | GenAI Summit 2025 — Austin, TX (October 8, 2025) Session Title: Opening Pandora’s Box: Building Effective Multimodal Feedback Loops Speaker: Denise Kutnick, Co-Founder & CEO, Variata Talk Track: Multimodal Systems in Production Abstract: AI market maps are overflowing with multimodal SDKs claiming to unify vision, language, audio, and more — but when they fail in production, debugging them often feels like being trapped in a black box. In this session, Denise Kutnick opens that box and walks through how to design multimodal feedback loops that make complex AI systems transparent, reliable, and interpretable in production environments. You’ll learn how to move from closed-box to open-box workflows, exposing intermediate signals within agentic pipelines to gain fine-grained control, faster debugging, and better calibration to user needs. Denise also shares how to define human-understandable evaluation checkpoints and build data pipelines using declarative tools, computed columns, and batch execution—all to improve feedback loops without slowing down deployment. What you’ll learn: • The difference between closed-box and open-box multimodal workflows • How to build interpretable feedback loops for real-world AI systems • Why defining human-centered evaluation checkpoints is essential • How to design data pipelines that enhance visibility and reduce risk in production • How proper scaffolding gives control and interpretability regardless of SDK choice.
