Why FeatureOps Might Be the Future of Software Delivery

I invited Egil Østhus to unpack a simple idea that tends to get lost in release day pressure. DevOps gets code to production quickly, but users experience features, not pipelines. Egil is the founder of Unleash, an open source feature management platform with close to 30 million downloads, and he argues that the next step is FeatureOps. It is a mindset and a set of practices that separate deployment from release, so teams can place code in production, light it up for a small cohort, learn, and only then scale out with confidence. Here is the thing. Controlled rollouts, clear telemetry, and fast rollback reduce risk without slowing teams down. Egil explains how FeatureOps connects engineering effort to business outcomes through gradual exposure, full stack experimentation, and what he calls surgical rollback. Instead of ripping out an entire release when one part misbehaves, teams can disable the offending capability and keep the rest of the value in place. It sounds straightforward because it is, and that is the point. Less drama, more learning, better results. We also talk about culture. When releases repeatedly disappoint, trust between product and engineering frays. Egil shares examples where Unleash helped a hardware and software company move from blame to shared ownership by making rollout plans visible and collaborative. Another client, an ERP vendor, discovered that early feedback from a small group of users allowed them to ship a leaner version that met the need without months of extra scope. That is how FeatureOps saves money and tempers expectations while still delighting customers. AI enters the story too. Code is shipping faster, but reliability can wobble when autogenerated changes move through pipelines. Egil sees feature management as a practical control plane for this new reality. Feature flags provide a real time safety net and, if needed, a kill switch for AI powered functionality. Teams can keep experimenting while protecting users and brand equity. If you want to move beyond release day roulette, this episode offers a practical playbook. We cover privacy first design, open source flexibility, and why metadata from FeatureOps will help leaders study how their organizations truly build. To learn more, visit getunleash.io or search for Unleash in your favorite tool, then tell me how you plan to measure your next rollout’s impact. ********* Visit the Sponsor of Tech Talks Network: Land your first job  in tech in 6 months as a Software QA Engineering Bootcamp with Careerist https://crst.co/OGCLA      

I invited Atalia Horenshtien to unpack a topic many leaders are wrestling with right now. Everyone is talking about AI agents, yet most teams are still living with rule based bots, brittle scripts, and a fair bit of anxiety about handing decisions to software. Atalia has lived through the full arc, from early machine learning and automated pipelines to today’s agent frameworks inside large enterprises. She is an AI and data strategist, a former data scientist and software engineer, and has just joined Hakoda, an IBM company, to help global brands move from experiments to outcomes. The timing matters. She starts on the 18th, and this conversation captures how she thinks about responsible progress at exactly the moment she steps into that new role. Here’s the thing. Words like autonomy sound glamorous until an agent faces a messy real world task. Atalia draws a clear line between scripted bots and agents with goals, memory, and the ability to learn from feedback. Her advice is refreshingly grounded. Start internal where you can observe behavior. Put human in the loop review where it counts. Use role based access rather than feeding an LLM everything you own. Build an observability layer so you can see what the model did, why it did it, and what it cost. We also get into measurements that matter. Time saved, cycle time reduction, adoption, before and after comparisons, and a sober look at LLM costs against any reduction in FTE hours. She shares how custom cost tracking for agents prevents surprises, and why version one should ship even if it is imperfect. Culture shows up as a recurring theme. Leaders need to talk openly about reskilling, coach managers through change, and invite teams to be co creators. Her story about Hakoda’s internal AI Lab is a good example. What began as an engineer’s idea for ETL schema matching grew into agent powered tools that won a CIO 100 award and now help deliver faster, better outcomes for clients. There are lighter moments too. Atalia explains how she taught an ex NFL player the basics of time series forecasting using football tactics. Then she takes us behind the scenes with McLaren Racing, where data and strategy collide on the F1 circuit, and admits she has become a committed fan because of that work. If you want a practical playbook for moving from shiny demos to dependable agents, this episode will help you think clearly about scope, safeguards, and speed. Connect with Atalia on LinkedIn, explore Hakoda’s work at hakoda.io, and then tell me how you plan to measure your first agent’s value. ********* Visit the Sponsor of Tech Talks Network: Land your first job  in tech in 6 months as a Software QA Engineering Bootcamp with Careerist https://crst.co/OGCLA  

Here’s the thing. Connecting thousands of devices is the easy part. Keeping them resilient and secure as you grow is where the real work lives. In this episode, I sit down with Iain Davidson, Senior Product Manager at Wireless Logic, to unpack what happens when connectivity, security, and operations meet in the real world. Wireless Logic connects a new IoT device every 18 seconds, with more than 18 million active subscriptions across 165 countries and partnerships with over 750 mobile networks. That reach brings hard lessons about where projects stall, where breaches begin, and how to build systems that can take a hit without taking your business offline. Iain lays out a simple idea that more teams need to hear. Resilience and security have to scale at the same pace as your device rollouts. He explains why fallback connectivity, private networking, and an IoT-optimised mobile core such as Conexa set the ground rules, but the real differentiator is visibility. If you cannot see what your fleet is doing in near real time, you are guessing. We talk through Wireless Logic’s agentless anomaly and threat detection that runs in the mobile core, creating behavioural baselines and flagging malware events, backdoors, and suspicious traffic before small issues become outages. It is an early warning layer for fleets that often live beyond the traditional IT perimeter. We also get honest about risk. Iain shares why one in three breaches now involve an IoT device and why detection can still take months. Ransomware demands grab headlines, but the quiet damage shows up in recovery costs, truck rolls, and trust lost with customers. Then there is compliance. With new rules tightening in Europe and beyond, scaling without protection does not only invite attackers. It can keep you out of the market. Iain’s message is clear. Bake security in from day one through defend, detect, react practices, supply chain checks, secure boot and firmware integrity, OTA updates, and the discipline to rehearse incident playbooks so people know what to do when alarms sound. What if you already shipped devices without all of that in place? We cover that too. From migrating SIMs into secure private networks to quarantining suspect endpoints and turning on core-level detection without adding agents, there are practical ways to raise your posture without ripping and replacing hardware. Automation helps, especially at global scale, but people still make the judgment calls. Train your teams, run simulations, and give both humans and digital systems clear rules for when to block, when to escalate, and when to restore from backup. I left this conversation with a simple takeaway. Growth is only real if it is durable. If you are rolling out EV chargers, medical devices, cameras, industrial sensors, or anything that talks to the network, this episode gives you a working playbook for scaling with confidence. Connect with Iain on LinkedIn, explore the IoT security resources at WirelessLogic.com, or reach the team at [email protected]. ********* Visit the Sponsor of Tech Talks Network: Land your first job  in tech in 6 months as a Software QA Engineering Bootcamp with Careerist https://crst.co/OGCLA

We talk a lot about AI as if it can fix broken systems. But what happens when the underlying data is too messy, too slow, or too disconnected to support anything useful? That’s the problem Manish Sood, founder and CTO of Reltio, has spent the last decade working to solve. Reltio is not your average data company. It sits behind some of the world’s most recognisable brands, helping names like L’Oréal, Pfizer, HP, and CarMax modernise how they manage and activate data across the business. What they all share is a recognition that outdated systems and disconnected records don’t just slow down insights. They actively block innovation. Manish breaks this down with uncommon clarity. He calls it “data debt”—the invisible burden of stale, incomplete, or fragmented information that quietly kills speed and adds risk. It’s not just a technical problem. It’s a leadership challenge, especially as businesses adopt generative AI tools that rely on clean, contextual data to function reliably. We explore how real-time intelligence is changing the way companies operate across customer experience, fraud detection, and supply chain resilience. Manish shares examples from enterprise clients who have moved from legacy systems to unified platforms, and how that shift enabled smarter decision-making at scale. From personalised retail offers to proactive healthcare outreach, the stories point to one common truth: if the data isn’t trusted, the AI cannot be either. There’s also a new role emerging inside many companies—the data steward as AI enabler. These are the people ensuring that data isn’t just stored, but shaped. Human-guided, explainable, traceable. That clarity is key to responsible AI, especially in sectors where compliance and reputation are tightly linked. Manish also explains how Reltio’s platform helps businesses protect against AI vulnerabilities by enabling resilient data pipelines, consistent governance, and real-time monitoring. In a world where data is created and used simultaneously, batch syncing is not enough. Real-time pipelines give companies the confidence to experiment with AI without falling into chaos. If your business is chasing innovation without cleaning up its data layer first, this conversation is a wake-up call. Manish shows that the future of AI isn’t about who builds the best model. It’s about who feeds it the best foundation. ********* Visit the Sponsor of Tech Talks Network: Land your first job  in tech in 6 months as a Software QA Engineering Bootcamp with Careerist https://crst.co/OGCLA    

Here's the thing. Most enterprise AI pitches talk about scale and speed. Fewer talk about trust, tone, and culture. In this conversation with Inflection AI's Amit Manjhi and Shruti Prakash, I explore a different path for enterprise AI, one that combines emotional intelligence with analytical horsepower, enabling teams to ask more informed questions of their data and receive answers that are grounded in context. Amit's story sets the pace. He is a three-time founder, a YC alum, and a CS PhD who has solved complex problems across mobile, ad tech, and data. Shruti complements that arc with a product lens shaped by real operational trenches, from clean rooms to grocery retail analytics.  Together, they built BoostKPI during the pandemic, transforming natural language into actionable insights, and then joined Inflection AI to help refocus the company on achieving enterprise outcomes. Their shared north star is simple to say yet tricky to execute. Make data analysis conversational, accurate, and emotionally aware so people actually use it. We unpack Inflection's shift from Pi's consumer roots to privacy-first enterprise tools. That history matters because it gives the team a head start on EQ. When you combine a deep well of human-to-AI conversations with modern LLMs, you get systems that explain, probe, and adapt rather than dump charts and call it a day.  Shruti breaks down what dialogue with data looks like in practice. Think back-and-forth exchanges that move from "what happened" to "why it happened," then on to "where else this pattern appears" and "what to do next," all grounded in an organization's language and values. Amit takes us under the hood on deployment choices and ownership. If a customer wants on-prem or VPC, they get it. If they're going to fine-tune models to their vernacular, they can. The model, the insights, and the guardrails remain in the customer's control. I enjoyed the honesty around adoption. Chasing AGI makes headlines, but it rarely helps a merchandising manager spot an early drop in lifetime value or a CX lead understand churn risk before quarter end. The duo keeps the conversation grounded in everyday questions that drive numbers and reduce meetings. They describe a path where EQ and IQ come together to form what Shruti calls contextual intelligence, and where brands can trust AI agents to assist without losing ownership or voice. If you care about making data useful to more people, and you want AI that sounds like your company rather than a generic assistant, this one is for you. We cover startup lessons, the reality of cofounding as a couple during lockdowns, and how Inflection is working with large enterprises to bring conversational analysis to real workloads. It is a grounded look at where enterprise AI is heading, and a timely reminder that technology should elevate humans, not replace them. ********* Visit the Sponsor of Tech Talks Network: Land your first job  in tech in 6 months as a Software QA Engineering Bootcamp with Careerist https://crst.co/OGCLA

Here’s the thing. Payments only look simple from the outside. In this Tech Talks Daily episode, I sit down with Roberto “Reks” Kafati, CEO and co-founder of DEUNA, to unpack how a scrappy one-click checkout idea grew into an intelligent payments infrastructure that now touches a large slice of Mexico’s online economy.  Reks explains why Latin America’s high decline rates aren’t just an inconvenience but a growth killer, and how DEUNA’s early focus on orchestration and checkout opened the door to something bigger. When a region routinely sees more than four out of ten online transactions knocked back, the bar for reliability sits in a different place. That practical problem set the stage for what came next. Athia, Real-Time Decisions, and 638 Signals per Transaction DEUNA’s pivot point came when merchants asked a fair question. With all this payment data flying through the system, what should we do with it? The answer is Athia, DEUNA’s AI-powered layer that watches every transaction and feeds merchants real-time insight, routing choices, and suggested actions. It is not another dashboard you promise to check and then ignore by Friday. It is a reasoning engine that sits on top of 638 data points per transaction and turns mess into movement. That is how you recover revenue without punishing good customers with extra friction, how you avoid surprise fees from networks, and how you protect recurring revenue when a processor wobbles. Reks walks us through results that speak plainly. Ramped merchants saw conversion lift from the original one-click experience. The infrastructure tier recovers meaningful GMV and trims fees. Enterprise clients report double-digit ROI and stick around for the compounding effect. Building Through Adversity and Betting on the Right Layer What resonated most was the human story behind the metrics. DEUNA was born in the first months of the pandemic, shaped by the shock that hit real-world businesses when revenue fell off a cliff and marketplaces became a lifeline with strings attached. Reks shares an unvarnished look at a tough 2023, the kind of year founders rarely talk about on record. Revenues dipped, deals went sideways, life got complicated. The team chose resilience and doubled down on a two-year vision. That bet is paying off. Over the past twenty-four months the company has grown at a pace that would bend a chart, and the focus has shifted from commoditizing orchestration to productizing intelligence. Put simply, earn trust at checkout, then make the data work for the merchant in real time. Agentic Commerce, US Expansion, and What Comes Next We also look forward. If chat interfaces begin to mediate more buying decisions, merchants will need infrastructure that can think, not just connect endpoints. That is the territory DEUNA calls intelligent infrastructure, and it is where Athia operates every day.  The company is now in active conversations with major US retailers, confident after winning head-to-head enterprise evaluations. Reks frames the opportunity without hype. If you can see acceptance trends by processor, by country, by card type, and act in the moment, you keep customers, protect margins, and avoid death by a thousand false declines. If you cannot, competitors will gladly welcome your frustrated shoppers. If you care about the real mechanics of growth, this conversation is for you. We talk conversion lift, recovered revenue, and the gritty bits of building a payments company that merchants actually rely on. We also talk about the days that test your resolve and the tenth day that reminds you why you started. ********* Visit the Sponsor of Tech Talks Network: Land your first job  in tech in 6 months as a Software QA Engineering Bootcamp with Careerist https://crst.co/OGCLA

I sat down with Leo de Araujo, Head of Global Business Innovation at Syntax Systems, to unpack a problem every SAP team knows too well. Years of enhancements and quick fixes leave you with custom code that nobody wants to document, a maze of SharePoint folders, and hard questions whenever S/4HANA comes up. What does this program do. What breaks if we change that field. Do we have three versions of the same thing. Leo’s answer is Syntax AI CodeGenie, an agentic AI solution with a built-in chatbot that finally treats documentation and code understanding as a living part of the system, not an afterthought. Here’s the thing. CodeGenie automates the creation and upkeep of custom code documentation, then lets you ask plain-language questions about function and business value. Instead of hunting through 40-page PDFs, teams can ask, “Do we already upload sales orders from Excel,” or “What depends on this BAdI,” and get an instant explanation. That changes migration planning. You can see what to keep, what to retire, and where standard capabilities or new extensions make more sense, which shortens the path to S/4HANA Cloud and helps you stay on a clean core. We also talk about how this is delivered. CodeGenie runs on SAP Business Technology Platform, connects through standard APIs, and avoids intrusive add-ons. It is compatible with SAP S/4HANA, S/4HANA Cloud Private Edition through RISE with SAP, and on-premises ECC. Security comes first, with tenant isolation for each customer and no custom code shared externally or used for AI model training. The result is a setup that respects enterprise guardrails while still giving developers and architects fast answers. Clean core gets a plain explanation in this episode. Build outside the application with published APIs, keep upgrades predictable, and innovate at the edge where you can move quickly. CodeGenie gives you the visibility to make that real, surfacing what you actually run today and how it ties to outcomes, so you can design a migration roadmap that fits the business rather than guessing from stale documents. Leo also previews the Gen AI Starter Pack, launching September 9. It bundles a managed, model-flexible platform with workshops, use-case ideation, and initial builds, so teams can move from curiosity to working solutions without locking themselves into a single provider. Paired with CodeGenie and Syntax’s development accelerators, the Starter Pack points toward something SAP leaders have wanted for years, a practical way to shift from in-core customizations to clean-core extensions with much less friction. If you are planning S/4HANA, balancing hybrid and multi-cloud realities, or simply tired of tribal knowledge around critical programs, this conversation is for you. We get specific about how CodeGenie works, where it saves time and cost, and how Syntax is shaping a playbook for AI that helps teams deliver results they can trust. ********* Visit the Sponsor of Tech Talks Network: Land your first job  in tech in 6 months as a Software QA Engineering Bootcamp with Careerist https://crst.co/OGCLA

Drew Allen, CEO of Grace Technologies, shares real stories from the floor, the ideas shaping safer plants, and why culture matters more than slogans. Drew’s background stretches from a family line linked to Samuel Morse to teenage years in China to global business development at 3M. That range shows up in how he leads. He listens, he moves fast, and he expects teams to work on things that matter. In his world that means saving electricians from shocks and arc flash while helping manufacturers modernize without losing their soul. Grace started with mechanical and analog products, then took the hard road into fully digital systems. The shift took time and patience. Today their platform brings sensors, AI, and cloud tooling into maintenance and safety. The example that stuck with me is a proximity band for electricians. It lights, beeps, and vibrates as a worker approaches live voltage. At TriCity, that band prevented three near misses in a three month pilot. A fourth incident still ended in a hospital visit and a costly outage because the worker left the band in his car. Another apprentice nearly placed a hand on a live bus bar until the band told him something was wrong. These moments remind you that technology can change a day and a life. Drew’s take on culture is refreshingly direct. Values are not a poster. They are a filter for who you hire. He looks for customer obsession, ownership, curiosity, and candid communication. Then he pairs that with high expectations and real care. Autonomy comes with accountability. Impact matters. If someone does not want to work on meaningful problems, this is not their place. It sounds firm. It also explains why the company keeps earning top workplace recognition while raising the bar on performance. We also talked about Maple Studios, the startup incubator Drew launched in Davenport, Iowa. He sees gaps in the industrial ecosystem. Fewer big exits. Slow adoption cycles. Founders stuck inside large companies. Maple gives them tools, space, and hard feedback so they can iterate faster and build things factories will actually deploy. His advice is simple. Ship, learn, and repeat. Do customer reviews early. Expect a thousand small gotchas. Move through them rather than pretending they will not appear. Looking ahead, Drew expects robotics to accelerate for a very practical reason. Companies cannot find enough people. Dangerous work will be automated. He imagines maintenance tasks shifting toward humanoid robots, with machines designed so robotic agents can service them. He also references GM’s self healing language to point at a coming blend of sensing, prediction, and automated repair.  On AI, he shares Satya Nadella’s challenge. Measure productivity and GDP impact rather than hype. The promise is there. The scoreboard will tell the story. If you work in industrial tech, this conversation lands close to home. You will hear how to bring digital tools into legacy environments, how to design for safety from the start, and how to keep teams motivated without losing kindness. You will also catch an open invitation. Drew wants to partner with builders who care about this space. If that is you, reach out to him on LinkedIn or visit graceport.com. And if you are curious about the band that vibrates before a bad day begins, this episode is a good place to start. ********* Visit the Sponsor of Tech Talks Network: Land your first job  in tech in 6 months as a Software QA Engineering Bootcamp with Careerist https://crst.co/OGCLA

I spoke with Kalai Ramea at a timely moment. We recorded this conversation during a heatwave in the UK, which made her work at Planette AI feel very real. Kalai calls herself an all-purpose scientist, with a path that runs through California climate policy, Xerox PARC, and now a startup focused on the forecast window that most people ignore. Not tomorrow’s weather. Not far-off climate scenarios. The space in between. Two weeks to two months out, where decisions get made and money is on the line. Kalai explains Planette AI’s idea of scientific AI in plain words. Instead of learning from yesterday’s weather patterns and hoping the future looks the same, their models learn physics from earth system simulations. Ocean meets atmosphere, energy moves, and the model learns those relationships directly. That matters in a warming world where history is a shaky guide. It also shortens time to insight. Traditional models can take weeks to run. If the output arrives after the risky period has passed, it is trivia.  tte AI is building for speed and usefulness. The value shows up in places you can picture. Event planners deciding whether to green-light a festival. Airlines shaping schedules and staffing. Farmers choosing when to plant and irrigate. Insurers pricing risk without leaning only on the past. Kalai shared a telling backcast of Bonnaroo in Tennessee, where flooding forced a last-minute cancellation. Their system showed heavy-rain signals weeks ahead. That kind of lead time changes outcomes, budgets, and stress levels. From Jargon To Decisions What I appreciate most about this story is the focus on access. Too many forecasts live in papers that only specialists read. Kalai and team are working to strip away jargon and deliver answers people can act on. Will it rain enough to trigger a payout. Will a heat threshold be crossed. Will the next month bring the kind of wind that matters for grid operations. The delivery matters as much as the math. NetCDF files might work for researchers, but a map, a simple number, or a chat interface is what users reach for when time is short. There is also a financial thread running through this work. Climate risk now shapes crop insurance, carbon programs, and balance sheets. Parametric insurance is growing because it is simple. Set a threshold. If it hits, the policy pays. Better medium-range signals make those products fairer and more useful. Kalai describes Planette AI’s role as a baseline layer others can build on, a kind of AWS for climate intelligence. That framing fits. No single company will build every app in this space. A reliable core makes the rest possible. Kalai’s path ties it all together. Policy taught her how decisions get made. PARC sharpened her instincts for practical AI. PlanetteAI is the result. If you care about planning beyond next week, this episode will give you a new way to think about forecasts and the tools that power them. I will add the blog link Kalai shared in the show notes. In the meantime, if you are in agriculture, travel, energy, or insurance, ask yourself a simple question. What would you change if you had a trustworthy signal three to eight weeks ahead. ********* Visit the Sponsor of Tech Talks Network: Land your first job  in tech in 6 months as a Software QA Engineering Bootcamp with Careerist https://crst.co/OGCLA

During the IT Press Tour, I had the pleasure of speaking with Weimo Liu, CEO and co-founder of PuppyGraph, and hearing firsthand how his team is rethinking graph technology for the enterprise. In this episode of Tech Talks Daily, Weimo joins me to share the story behind PuppyGraph’s “zero ETL” approach, which lets organizations query their existing data as a graph without ever moving or duplicating it. We discuss why graph databases, despite their promise, have struggled with mainstream adoption, often because of complex pipelines and heavy infrastructure requirements. Weimo explains how PuppyGraph borrows from his time at TigerGraph and Google’s F1 engine to build something new: a distributed query engine that maps tables into a logical graph and delivers subsecond performance on massive datasets. That shift opens the door for use cases in cybersecurity, fraud detection, and AI-driven applications where latency and accuracy matter most. We also unpack the developer experience. Instead of rewriting schemas or reloading data every time requirements change, PuppyGraph allows teams to define nodes and edges directly from existing tables. That design lowers the barrier for SQL-focused teams and accelerates time to value. Weimo even touches on the role of graph in reducing AI hallucinations, showing how structured relationships can make enterprise AI systems more reliable. What struck me most in our conversation is how PuppyGraph’s playful branding belies its serious engineering depth. Behind the “puppy” name lies a distributed engine built to scale with today’s data volumes, backed by strong early adoption and a team that listens closely to customer needs. Whether you’re exploring graph for cybersecurity, AI chatbots, or supply chain analytics, this discussion offers a glimpse of how the next generation of graph tech might finally break free from its niche and go mainstream. ********* Visit the Sponsor of Tech Talks Network: Land your first job  in tech in 6 months as a Software QA Engineering Bootcamp with Careerist https://crst.co/OGCLA