MLOps.community

Kris Beevers, CEO and co-founder of NetBox Labs and veteran network engineer with a Ph.D., discusses how modern AI datacenters handle hypergrowth. He talks about modeling infrastructure as a single system of record, tackling power and procurement bottlenecks, and using programmatic blueprints and digital twins to speed builds. He highlights rapid iteration, robotics in racking, and the need for vendor data standards for automation.

Mike Oaten, founder and CEO of TIKOS, builds AI assurance and explainability tools for high-stakes systems. He discusses real-time neuron monitoring, capturing internal activations and causality traces, and translating fuzzy regulations into concrete tests. Conversations cover regulatory risks of closed models, creating golden profiles for gating, and mapping internal traces to audit-ready explainability.

Paulo Vasconcellos, Principal Data Scientist for Generative AI Products at Hotmart and co-founder of Data Hackers, builds AI tools for creators and learners. He discusses using LLMs alongside classic NLP for speed and cost. He explores multilingual model choices, agent-as-a-product creator tools, Hotmart Tutor as a 24/7 teacher, and engineering tradeoffs for scalable ML and safety guardrails.

Wilder Lopes, a second-time founder and CEO of Ogre.run, dives into the pressing challenges of AI infrastructure. He discusses how many workloads can effectively run on idle CPUs instead of GPUs, addressing the often overlooked memory bottleneck. Wilder highlights innovations like CXL memory expansion and the need for better developer tooling for non-GPU hardware. He envisions a future with a diverse 'NeoCloud', emphasizing the importance of equipping developers with hardware knowledge and leveraging second-hand data centers for global AI advancements.

Corey Zumar, a Product Manager at Databricks and MLflow maintainer, joins Jules Damji, a Lead Developer Advocate and expert on Spark, and Danny Chiao, an Engineering Leader focused on AI governance. They discuss how MLflow is evolving to support generative AI and tool-calling agents. Key topics include challenges in multi-turn conversations, the importance of feedback for optimization, and the need for unified platforms to streamline AI workflows. They also delve into data governance strategies and the balance of cost and quality in model management.

Euro Beinat, Global Head of AI at Prosus Group, and Mert Öztekin, Chief Technology Officer at Just Eat Takeaway.com, dive into the transformative role of AI in tech leadership. They discuss how AI shifts the CTO's focus from feature delivery to organizational change. Broad access to AI tools can spark innovation, while measured rollouts build trust. The duo emphasizes collective experimentation and the need for practical governance to ensure safe AI use. Plus, they explore the impact of mini automations across departments and the importance of bold targets for engagement.

Zengyi Qin, founder of the OpenAGI Foundation, discusses groundbreaking technology enabling computers to think and act autonomously. He details innovative training methods using large-scale sandboxes and real-world applications, highlighting Lux, a model that outperforms competitors like Gemini. Zengyi explains the agent framework for efficient task management and envisions AI replacing traditional input methods in the near future. He emphasizes the importance of continuous model improvement to tackle performance challenges, paving the way for a new era of human-computer interaction.

Varant Zanoyan, Co-founder and CEO of Zipline AI, and Nikhil Simha Raprolu, Co-founder and CTO at Zipline AI, delve into the evolution of AI infrastructure. They share insights on the compute-first approach of Cronon that emerged from Airbnb, emphasizing real-time features over traditional storage models. The duo explains the complexities of orchestrating signals and pipelines, the challenges of point-in-time correctness, and the importance of governance. They also discuss how Cronon integrates embeddings and real-time workflows, reflecting on its open-sourcing journey with Stripe.

In this discussion, guests Chiara Caratelli and Alex Salazar dive into the intricacies of AI agent tooling. Chiara shares insights from her work at Prosus, focusing on UX and the challenges of clarity in tool definitions, highlighting how ambiguity can lead to edge cases. Alex contrasts traditional APIs with dynamic tools that capture agent intentions to minimize latency. Together, they explore the balance of tool usage, governance, and the future of multi-agent systems, advocating for a structured approach to ensure efficient production-ready AI agents.

Jonathan Wall, CEO of Runloop.ai and a former Google engineer, dives into the future of AI agents and the importance of sandboxed environments. He reveals how sandboxes create safe spaces for agents to operate while preventing security risks. Wall discusses building efficient agent infrastructures, the advantages of isolated compute environments, and the significance of creating Git-like workflows for enterprises. He also explores how refining agent performance through failed runs can drive iterative advancements, revolutionizing how agents interact with data and each other.