3368: Nerdio Builds a Remote-First Unicorn Focused on Azure Optimization

Here’s the thing. Most enterprise AI talk today starts with chatbots and ends with glossy demos. Meanwhile, the data that actually runs a business lives in rows, columns, and time stamps. That gap is where my conversation with Vanja Josifovski, CEO of Kumo AI really comes alive. Vanja has spent two and a half decades helping companies turn data into decisions, from research roles at Yahoo and Google to steering product and engineering at Pinterest through its IPO and later leading Airbnb Homes. He’s now building Kumo AI to answer an old question with a new approach: how do you get accurate, production-grade predictions from relational data without spending months crafting a bespoke model for each use case? Vanja explains why structured business data has been underserved for years. Images and text behave nicely compared to the messy reality of multiple tables, mixed data types, and event histories. Traditional teams anticipate a prediction need, then kick off a long feature engineering and modeling process. Kumo's Relational Foundation Model, or RFM, flips that script. Pre-trained on a large mix of public and synthetic data warehouses, it delivers task-agnostic, zero-shot predictions for problems like churn and fraud. That means you can ask the model questions directly of your data and get useful answers fast, then fine-tune for another 15 to 20 percent uplift when you’re ready to squeeze more from your full dataset. What stood out for me is how Kumo removes the grind of manual feature creation. Vanja draws a clear parallel to computer vision’s shift years ago, when teams stopped handcrafting edge detectors and started learning from raw pixels. By learning directly from raw tables, Kumo taps the entirety of the data rather than a bundle of human-crafted summaries. The payoff shows up in the numbers customers care about, with double-digit improvements against mature, well-defended baselines and the kind of time savings that change roadmaps. One customer built sixty models in two weeks, a job that would typically span a year or more. We also explore how this fits with the LLM moment. Vanja doesn’t position RFM as a replacement for language models. He frames it as a complement that fills an accuracy gap on tabular data where LLMs often drift. Think of RFM as part of an agentic toolbox: when an agent needs a reliable prediction from enterprise data, it can call Kumo instead of generating code, training a fresh model, or bluffing an answer. That design extends to the realities of production as well. Kumo’s fine-tuning and serving stack is built for high-QPS environments, the kind you see in recommendations and ad tech, where cost and latency matter. The training story is another thread you’ll hear in this episode. The team began with public datasets, then leaned into synthetic data to cover scenarios that are hard to source in the wild. Synthetic generation gives them better control over distribution shifts and edge cases, which speeds iteration and makes the foundation model more broadly capable upon arrival. If you care about measurable outcomes, this episode shows why CFOs pay attention when RFM lands. Vanja shares examples where a 20 to 30 percent lift translates into hundreds of thousands of additional monthly active users and direct revenue impact. That kind of improvement isn’t theory. It’s the difference between a model that nudges a metric and a model that moves it. By the end, you’ll have a clear picture of what Kumo AI is building, why relational data warrants its own foundation model, and how enterprises can move from wishful thinking to practical wins. Curious to try it yourself? Vanja also points to a sandbox where teams can load data and ask predictive questions within a notebook, then compare results against in-house models. If your AI plans keep stalling on tabular reality, this conversation offers a way forward that’s fast, accurate, and designed for the systems you already run.

When VMware Cloud Foundation 9.0 launched in June, it marked more than just another release. It was the clearest signal yet that Broadcom is betting big on the modern private cloud. In this episode of Tech Talks Daily, I sat down with Prashanth Shenoy, who leads marketing and learning for the VCF division at Broadcom, to discuss what the launch means for enterprises and how those themes are playing out live at VMware Explore in Las Vegas. Prashanth shares how VCF 9.0 was designed to help enterprises operate private clouds with the same simplicity and scale as public hyperscalers, while keeping sovereignty, security, and cost predictability front and center. He explains why this release is more than an infrastructure update. It’s a shift toward a workload-agnostic, developer-centric platform where virtual machines, containers, and AI workloads can run side by side with a consistent operational experience. We also unpack Broadcom’s headline announcements at the show. From making VCF an AI-native platform to embedding private AI services directly into the foundation, the message is clear: the AI pilots of the past are moving into production, and Broadcom wants VCF to be the default home for enterprise AI. Another major theme is cyber compliance at scale, with VCF now offering continuous enforcement, rapid ransomware recovery, and advanced security services that address today’s board-level concerns. But perhaps the biggest takeaway is the momentum. Nine of the top ten Fortune companies are now running on VCF, more than 100 million cores have been licensed, and dozens of enterprises—from global giants to mid-sized insurers—are on stage at VMware Explore sharing their adoption stories. The so-called “cloud reset” that Prashanth has written about is not just theory. Companies are rethinking their cloud strategies, seeking cost transparency, avoiding waste, and building resilient, AI-ready private clouds. This conversation highlights how Broadcom is doubling down on VCF with a singular focus, a massive R&D commitment, and a clear vision of where private cloud is headed. If you want to understand why private AI, developer services, and cyber resilience are now central to enterprise strategy, this is a conversation worth hearing.

At VMware Explore in Las Vegas, the buzz wasn’t just about generative AI, but about where and how it should run. My guest is Tasha Drew, Director of Engineering for the AI team in the VMware Cloud Foundation division at Broadcom, who has been at the center of this conversation. Fresh off the main stage, where she helped debut VMware’s new Private AI Services and Intelligent Assist for VMware Cloud Foundation, Tasha joins me to unpack what these announcements mean for enterprises grappling with privacy, cost, and integration challenges. Tasha explains why private AI is resonating so strongly in 2025, outlining the three pillars that define it: protecting sensitive intellectual property, managing regulated or high-value data, and ensuring role-based control of fine-tuned models. She shares how organizations often start their AI journey in the public cloud, but as experimentation turns to production, cost pressures, data compliance, and proximity to data drive them toward private AI. We also dive into VMware’s own evolution toward building an AI-native private cloud platform. Tasha highlights the journey from deep learning VMs and Jupyter notebooks to full AI platform services that empower IT teams to deliver models efficiently, save money, and accelerate deployment of retrieval-augmented generation (RAG) applications. She introduces Intelligent Assist for VMware Cloud Foundation, an AI-powered guide that helps teams navigate complex deployments with context-aware support and step-by-step instructions. Beyond the technology, Tasha reflects on the broader ecosystem shifts, from partnerships with NVIDIA and AMD to the role of Model Context Protocol (MCP) in breaking down integration barriers between enterprise systems. She believes MCP represents a turning point, enabling seamless workflows between platforms that historically lacked incentive to work together. This conversation captures a pivotal moment where private AI is moving from theory into enterprise adoption. For leaders weighing their next move, Tasha provides both the strategic framing and the technical insight to understand why private AI has become one of the most talked-about forces shaping enterprise IT today.

Factories don’t usually make headlines at tech conferences, but what Audi is doing inside its production labs is anything but ordinary. At VMware Explore in Las Vegas, I sat down with Dr. Henning Löser, Head of the Audi Production Lab, to talk about how the automaker is reinventing its factory floor with a software-first mindset. Henning leads a small team he jokingly calls “the nerds of production,” but their work is changing how cars are built. Instead of replacing entire lines for every new piece of technology, Audi has found a way to bring the speed and flexibility of IT into the world of industrial automation. The result is Edge Cloud 4 Production, a system that takes virtualization technology normally reserved for data centers and applies it directly to manufacturing. In our conversation, Henning explained why virtual PLCs may be one of the biggest breakthroughs yet. They look invisible to workers on the line but give maintenance teams new transparency and resilience. We explored how replacing thousands of industrial PCs with centralized, virtualized workloads not only reduces downtime but also cuts energy use and simplifies updates. And yes, we even discussed the day a beaver chewed through one of Audi’s fiber optic cables and how redundancy kept production running without a hitch. This episode is about more than smart factories. It’s about how an industry known for heavy machinery is learning to think like the cloud. From scalability and sustainability to predictive maintenance and AI-ready infrastructure, Audi is showing how the car of the future starts with the factory of the future. If you’ve ever wondered how emerging technologies like virtualization and private cloud are reshaping the shop floor, this is a story you’ll want to hear.

In this episode of Tech Talks Daily, I speak with Jane Ostler from Kantar, the world’s leading marketing data and analytics company, whose clients include Google, Diageo, AB InBev, Unilever, and Kraft Heinz. Jane brings clarity to a debate often clouded by headlines, explaining why AI should be seen as a creative sparring partner, not a rival. She outlines how Kantar is helping brands balance efficiency with inspiration, and why the best marketing in the years ahead will come from humans and machines working together. We explore Kantar’s research into how marketers really feel about AI adoption, uncovering why so many projects stall in pilot phase, and what steps can help teams move from experimentation to execution. Jane also discusses the importance of data quality as the foundation of effective AI, drawing comparisons to the early days of GDPR when oversight and governance first became front of mind. From Coca-Cola’s AI-assisted Christmas ads to predictive analytics that help brands allocate budgets with greater confidence, Jane shares examples of where AI is already shaping marketing in ways that might surprise you. She also highlights the importance of cultural nuance in AI-driven campaigns across 90-plus markets, and why transparency, explainability, and human oversight are vital for earning consumer trust. Whether you’re a CMO weighing AI strategy, a brand manager experimenting with new tools, or someone curious about how the biggest advertisers are reshaping their playbooks, this conversation with Jane Ostler offers both inspiration and practical guidance. It’s about rethinking AI not as the end of creativity, but as the beginning of a new partnership between data, machines, and human imagination.

The enterprise network is under pressure like never before. Hybrid environments, cloud migrations, edge deployments, and the sudden surge in AI workloads have made it increasingly difficult to keep application connectivity secure and reliable. The old model of device-by-device, rule-based network management can’t keep up with today’s hyperconnected, API-driven world. In this episode of Tech Talks Daily, I sit down with Kyle Wickert, Field Chief Technology Officer at AlgoSec, to discuss the future of network management in the age of platformization. With more than a decade at AlgoSec and years of hands-on experience working with some of the world’s largest enterprises, Kyle brings an unfiltered view of the challenges and opportunities that IT leaders are facing right now. We talk about why enterprises are rapidly shifting to platform-based models to simplify network security, but also why that strategy can start to break down when dealing with multi-vendor environments. Kyle explains the fragmentation across cloud, on-prem, and edge infrastructure that keeps CIOs awake at night, and why spreadsheets and manual change processes are still far too common in 2025. He also shares why visibility, intent-based policies, and policy automation are becoming non-negotiable in reducing risk and friction. Kyle doesn’t just talk theory. He shares a real-world case study of a European financial institution that automated policy provisioning across firewalls and cloud infrastructure, integrated it with CI/CD pipelines, and reduced its change rejection rate from 25% to 4%. It’s a compelling example of how the right approach to network management can deliver measurable improvements in agility, security, and business satisfaction.  

AI is rapidly becoming part of the healthcare system, powering everything from diagnostic tools and medical devices to patient monitoring and hospital operations. But while the potential is extraordinary, the risks are equally stark. Many hospitals are adopting AI without the safeguards needed to protect patient safety, leaving critical systems exposed to threats that most in the sector have never faced before. In this episode of Tech Talks Daily, I speak with Ty Greenhalgh, Healthcare Industry Principal at Claroty, about why healthcare’s AI rush could come at a dangerous cost if security does not keep pace. Ty explains how novel threats like adversarial prompts, model poisoning, and decision manipulation could compromise clinical systems in ways that are very different from traditional cyberattacks. These are not just theoretical scenarios. AI-driven misinformation or manipulated diagnostics could directly impact patient care. We explore why the first step for hospitals is building a clear AI asset inventory. Too many organizations are rolling out AI models without knowing where they are deployed, how they interact with other systems, or what risks they introduce. Ty draws parallels with the hasty adoption of electronic health records, which created unforeseen security gaps that still haunt the industry today. With regulatory frameworks like the UK’s AI Act and the EU’s AI regulation approaching, Ty stresses that hospitals cannot afford to wait for legislation. Immediate action is needed to implement risk frameworks, strengthen vendor accountability, and integrate real-time monitoring of AI alongside legacy devices. Only then can healthcare organizations gain the trust and resilience needed to safely embrace the benefits of AI. This is a timely conversation for leaders across healthcare and cybersecurity. The sector is on the edge of an AI revolution, but the choices made now will determine whether that revolution strengthens patient care or undermines it. You can learn more about Claroty’s approach to securing healthcare technology at claroty.com.

In November, Alex Adamopoulos, CEO of Emergn, joined me on Tech Talks Daily to talk about transformation fatigue and why so many well-intentioned change programs leave people drained rather than inspired. This time, he’s back with a sharper question: if traditional transformation is broken, what actually works? His answer is refreshingly direct. Product thinking is strategic thinking, and it belongs everywhere in the enterprise, not just in product teams. In our conversation, Alex explains why HR, finance, and even legal teams now need product strategy skills as much as engineers or designers. He introduces Praxis, Emergn’s newly launched platform that rebrands their long-standing VFQ approach and now embeds product thinking across entire organizations. With its AI-powered coach Stella, Praxis is designed to support continuous learning while helping teams make better day-to-day decisions. We also discuss why outcomes, not deliverables, have become the accurate measure of digital success. Alex likens it to leaders constantly returning to their boards like entrepreneurs on Shark Tank, demonstrating incremental value before securing the next round of support. This shift in accountability changes how teams plan, learn, and invest. Another essential thread is the link between burnout and broken transformation models. Alex recently co-authored a paper with Harvard professor Amy Edmondson on “Breaking the Failure Cycle,” and he shares how adopting a product mindset can help organizations move past fatigue by focusing on outcomes, embracing uncertainty, and avoiding the endless reinvention trap. Whether you’re in a global enterprise grappling with AI adoption or a smaller company rethinking strategy, this episode is a reminder that transformation is not a program but a continuous practice. Product thinking offers a practical path forward, one that makes strategy executable, measurable, and, most importantly, sustainable.

In this episode of Tech Talks Daily, I’m joined for the third time by Justin Banon, the founder of Boson Protocol. A lot has changed since his last appearance. What started as a bold idea to decentralize e-commerce has now evolved into an ambitious, AI-first infrastructure aiming to redefine how we buy, sell, and interact with value itself. Justin walks us through the evolution of Boson. It began as a system for peer-to-peer digital and physical commerce, aiming to remove intermediaries like Amazon from the process. The next step was Fermion, a protocol designed specifically for high-value assets such as luxury goods, fine art, and real estate. Now, Boson is launching the Metasystem, a full-scale framework designed for what Justin calls “agentic commerce,” where AI agents transact on behalf of users. We talk about what that future looks like. Imagine an AI that not only shops for you but negotiates, verifies, and settles transactions securely. Justin predicts that in just a few years, these agents will outnumber human buyers and sellers by orders of magnitude. Boson’s mission is to build the decentralized rails for that world while avoiding the centralization traps of today’s tech giants. One particularly fascinating moment is our discussion of the Dolce & Gabbana glass suit. Purchased during the last bull run, this million-dollar piece of digital-physical fashion was not just fractionalized through Fermion but transformed into an AI persona called Dolce Lorien. This character now leads a gamified community campaign, rewarding participants with fractional ownership. It’s luxury meets sci-fi, wrapped in Web3 narrative. We also dig into what decentralized infrastructure really means for legacy brands. With Fermion, companies can reclaim the secondary market, verify resales, reconnect with past buyers, and turn their customer base into a true community. This isn’t just resale with a twist. It’s a new type of programmable loyalty system. Justin shares how AI doesn’t just enable commerce. It also supports his own personal growth. He reveals how he uses tools like ChatGPT and Speechify to create custom audio courses for niche subjects, walking through the woods while absorbing AI-generated masterclass-level insights. For those tired of seeing tech platforms control value creation while users are left with little ownership, this episode offers a glimpse of a different future. One where AI works directly for you. One where commerce is flexible, open, and fair. One where the infrastructure is built to stay that way.

In this episode of Tech Talks Daily, Neil sits down with Sean Li, co-founder and CEO of Magic Labs, to explore the intersection of crypto wallets, artificial intelligence, and the future of autonomous finance. Sean shares how Magic Labs has already onboarded over 50 million crypto wallets by pioneering simple login methods using email and SMS. Now, with Magic Newton, Sean is pushing into new territory where AI agents could securely manage digital assets on our behalf. From AI "concierges" executing investment strategies to cryptographic policy engines enforcing trust and control, the vision is clear: a financial internet where humans set the intent and machines handle the execution. We discuss the challenges of today’s fragmented crypto experience, how smart wallets and AI could abstract away complexity, and why Sean believes everyone with an email address will eventually have multiple agents acting on their behalf. You'll also hear why he compares this shift to building a digital institution or constitution for autonomous finance. Whether you're a developer, investor, or simply crypto-curious, this episode offers a fascinating look at where Web3, AI, and programmable trust may be heading next.