#350 How to Make Hard Choices in AI with Atay Kozlovski, Researcher at the University of Zurich

AI agents are getting easier to build and connect. Open standards now let them discover each other, pass messages and use shared tools. But connecting agents is not the same as getting them to work toward the same goal, and research on multi-agent systems shows that misalignment grows with every agent you add. For professionals deploying agents, this shows up as handoffs that fail and results that don’t add up, even when each agent is right on its own. How do agents agree on what a task means? What does a team need in place before adding more agents? And how do you know the combined result is better? Vijoy Pandey is SVP and GM of Outshift by Cisco, the company’s incubation engine for emerging technologies such as agentic AI and quantum networking. He has spent about three decades building distributed infrastructure at Google, IBM and startups, holds a PhD in Computer Science from UC Davis, and has filed more than 80 patents. In the episode, Richie and Vijoy explore the cognitive revolution and what it means for agents, the Internet of Cognition, coordination failures in multi-agent systems, agent protocols and guardrails, task-based access control, tiny teams with tokens (T3), organizational memory and context leaders, measuring whether agents are actually better, emergent collective intelligence, and much more. Links Mentioned in the Show: AGNTCY: open-source agent connectivity project Model Context Protocol (MCP) and Agent2Agent protocol (A2A) Outshift: Internet of Cognition Coherent Cognition, Vijoy’s Substack Michael Tomasello: A Natural History of Human Thinking Michael Levin: talks and interviews Connect with Vijoy AI-Native Course: Intro to AI for Work Related Episode: AI Agents Are the New Shadow IT (And Your Governance Isn’t Ready) with Stijn Christiaens, CEO at Collibra

Career goals in data have shifted quickly. Data scientist was long the default target, but AI engineer has appeared from almost nowhere, and companies are moving training toward AI literacy, agents, and prompt engineering. Certifications are becoming a common way to prove these skills, while many people wonder whether learning to code still matters when AI can write it for them. What are people choosing to learn right now? Which fundamentals still hold up? How do you prepare for a certification exam? And is the fear of AI replacing technical roles showing up in what people actually study? Pieter Moors is the Lead Data Scientist at DataCamp, where he works with the company's learner data to understand how people learn and how that behavior changes over time. He is part of DataCamp's new Data Insights program, which shares findings from that data with the world every couple of weeks. In the episode, Richie and Pieter explore what DataCamp's own learner data reveals, when enterprise and individual learners study, how the World Cup disrupted study time, why short regular sessions and streaks build habits, how AI tutors relate to certification pass rates, the shift from data scientist to AI engineer, and much more. Links Mentioned in the Show: Connect with Pieter DataCamp AI Tutor DataCamp Certifications Introduction to Python course DataCamp AI Assistant DataCamp mobile app DataCamp subreddit AI-Native Course: Intro to AI for Work Related Episode: The Best Moments from RADAR 11x: AI, Jobs, Agents and Skills

Most hiring decisions still come down to gut feel, and gut feel is full of blind spots — a favorite school, a firm handshake, the type of coffee someone carries into the room. As AI reshapes recruiting, the industry is being forced to ask whether decisions can be made on evidence instead of instinct. That means defining what actually predicts success in a role, testing every candidate against the same criteria, and being able to show the reasoning behind a hiring call rather than just a hunch. For anyone building a team, the question isn't whether bias exists — it's whether your process is built to catch it before it shapes who gets hired. Daniel Chait is the co-founder and CEO of Greenhouse, the hiring software used by tens of thousands of companies to run structured, data-driven hiring processes. Ophir Samson is Greenhouse's Head of Voice AI, leading the team following Greenhouse's acquisition of Ezra AI Labs, the voice AI interviewing company he founded. In the episode, Richie, Daniel, and Ophir explore the flood of AI-generated resumes drowning both job seekers and recruiters, how Voice AI interviews let candidates showcase themselves beyond a resume, structured hiring and scorecards for fairer decisions, bias audits and anti-cheat safeguards, EU AI Act compliance, and measuring success through pass-through rates and time savings, and much more. Links Mentioned in the Show: • Zapier <> Greenhouse: Conversational Application Voice AI • Paradigm Strategy • Anthropic • Pipecat on GitHub • Daily • Are Emily and Greg More Employable Than Lakisha and Jamal? (AEA) • Connect with Daniel: LinkedIn • Connect with Ophir: LinkedIn • AI-Native Course: Intro to AI for Work • Related Episode: How to Have a Machine Learning Career in 2026 New to DataCamp? Learn on the go using the DataCamp mobile app Empower your business with world-class data and AI skills with DataCamp for business

A job title can stay the same while the work underneath it changes completely. That idea ran through RADAR 11x, our online conference, and this recap keeps the best moments from four of the day's eight sessions. You will hear how to sort your tasks between you and AI, which durable skills hold their value, why pilots stall before production, and how software teams work with coding agents. Missed RADAR? Start here. Which tasks should AI take over? Which skills deserve your next hour of learning? Featured in this recap: Ben Zweig, CEO at Revelio Labs; Nicole Immorlica, Professor at Yale University and Senior Principal Researcher at Microsoft; Margaret Beier, Professor at Rice University; Matt Jones, EVP of Strategy at Cielo Talent; Krishnan Hariharan, CTO and VP of Engineering for Forge at Honeywell; Satesh Sonti, Principal Specialist Solutions Architect at AWS; Laurent Gil, President and co-founder at Cast AI; Joao Moura, CEO at CrewAI; Ali Arsanjani, Senior Director of Applied AI Engineering at Google; and DataCamp co-founders Jonathan Cornelissen (CEO) and Martijn Theuwissen (COO). In the episode, Richie rounds up the best moments from RADAR 11x, exploring whether AI will take your job, the durable skills worth building, why AI pilots stall, cutting token costs, how software teams work with agents, the future of the data analyst role, and much more. Links Mentioned in the Show: • We Are Not Machines (Sarah O'Connor) • Jevons paradox • AI-Driven Development Life Cycle (AI-DLC) (AWS) • The State of AI Careers 2026 (DataCamp) • Watch on demand: Humans On The Loop • Watch on demand: Don't Waste Your Time on AI Pilots • Watch on demand: The New Developer Workflow • Watch on demand: Closing and AMA • RADAR 11x event page • AI-Native Course: Intro to AI for Work • Related Episode: Data-Driven Workforce Analytics with Ben Zweig, CEO at Revelio Labs New to DataCamp? Learn on the go using the DataCamp mobile app Empower your business with world-class data and AI skills with DataCamp for business

AI agents are moving from demos into production, but many teams run into the same wall: responses that take too long and cost too much. As models grow bigger and reason through more steps before answering, every chained agent adds delay, and that adds up fast at scale. For data and AI professionals, this raises a practical question: how do you pick a model and an inference platform that keeps a multi-agent workflow fast without blowing the budget? And underneath that choice sits a bigger one — how much of the AI stack, from chips to software, should a company actually control? Sumti Jairath is Chief Architect at SambaNova Systems, where he's focused on machine learning, big data analytics, and software-defined hardware since 2017. Before that, he spent close to eight years as a Senior Hardware Architect at Oracle working on hardware acceleration for machine learning and data analytics, and earlier held processor and server design roles at Hewlett Packard and Sun Microsystems. He holds a B.Tech in Electronics and Communication Engineering from the National Institute of Technology, Kurukshetra. In the episode, Richie and Sumti explore why AI agents are slow and expensive, the architecture behind faster inference, choosing the right model and platform for a task, open-source models and sovereign AI, the full AI infrastructure stack, power efficiency, the AI data center debate, career paths in AI infrastructure, and much more. Links Mentioned in the Show: • NVIDIA's Groq acquisition, announced around GTC • Chris Olah, Anthropic • Anthropic and OpenAI's coding-agent-driven revenue growth • Ollama • Apple Mac mini • SambaNova careers • Connect with Sumti • AI-Native Course: Intro to AI for Work • Related Episode: From City Sewers to Sovereign AI with Russ Wilcox, CEO at ArtifexAI New to DataCamp? Learn on the go using the DataCamp mobile app Empower your business with world-class data and AI skills with DataCamp for business