Machine Minds

Episode 104 | AI & Automation Unlocking Materials Discovery | Joseph KrauseIn this episode of Machine Minds, I sit down with Joseph Krause, co-founder and CEO of Radical AI.We talk about Joseph’s path from Rice University and the Army National Guard to building Radical AI. He explains how AI and self-driving labs can speed up materials discovery, why failed experiments are critical data, and where this will first change the world—across aerospace, defense, semiconductors, and energy.Key topics:• The founding story of Radical AI• How AI + autonomous labs form a “materials flywheel”• Why negative results matter in science• Culture and hiring at Radical AI• Working with the White House on the 2025 National AI & R&D Strategy• Where early real-world impact will show upLinks:• To learn more about Radical AI, please visit https://www.radical-ai.com/• Radical AI is hiring! View open roles at: https://jobs.lever.co/RadicalAI• Connect with Joseph on LinkedIn: https://www.linkedin.com/in/josephfkrause/• Follow Joseph on X: https://x.com/josephfkrause?lang=enConnect with me:Greg Toroosian, Samson RoseLinkedIn: https://www.linkedin.com/in/gregtoroosian/

I sat down with Nick Snoad, VP at Marlinspike, to talk about the surge of capital into defense and dual-use tech—and how he actually evaluates teams, markets, and milestones at the earliest stages.What we get into:- Why dual-use matters: bigger markets, more paths to scale- How stage changes the bar: from “prove the tech and market” at seed to “execute and expand” post-A- Using non-dilutive routes (SBIRs, DIU, OTAs, TACFI/STRATFI) without losing the product roadmap- Where adoption is slow and where it’s breaking open: autonomy, simulation-driven wargaming, and the factory OS- Sectors on his radar now: energy (incl. nuclear and geothermal) and data-enabled manufacturing- What “value-add” looks like in defense VC (it’s not just a check)- Team signals he trusts: balanced founding groups, clear hiring plans, self-awareness on when to bring in a scaler- Advice to vets breaking in: go learn inside a startup, then decide if investing is your lanehttps://marlinspike.substack.com/ https://marlinspike.us/ Connect with Nick: https://www.linkedin.com/in/nick-snoad-340389157/ Connect with me: https://www.linkedin.com/in/gregtoroosian/

How do you get AI to seek the right data in the real world instead of drowning in all of it? In this episode, I sit down with Tara Javidi (UCSD professor and AI researcher) and Sam Bigdeli (repeat founder & former semiconductor supply‑chain exec), co-founders of Kav AI, to talk about “active physical intelligence”—hypothesis‑driven, curiosity‑led AI that hunts for the signals that matter in physical systems.We cover:Why passive, data-soaks-everything AI hits a wall in the physical worldHypothesis-driven learning: letting models ask “what should I look at next?”From oil & gas spills to structural failures—predicting the next “leak” like a language model predicts the next wordHandling massive, messy, multimodal sensor streams in real time (volume of context, not just length)Interpretability when your model is deciding which sensor to query and whyWhat academia gets wrong (and right) about startups—and vice versaThe hardest part of moving from novelty-driven research to problem-driven productHow (and when) to disagree productively as co-foundersLinks mentionedWebsite: www.kavai.comCompany LinkedIn: https://www.linkedin.com/company/kav-artificial-intelligence/Sam Bigdeli: https://www.linkedin.com/in/sam-bigdeli-5310b923/Tara Javidi: https://www.linkedin.com/in/tara-javidi-28b450155/🎙 Connect with meLinkedIn – https://www.linkedin.com/in/greg-toroosian

In this episode I sit down with Akshay Chalana, CEO & co‑founder of Saphira AI, to unpack how "TurboTax for robot compliance” can slash the time, cost, and confusion that stall hardware roll‑outs.We cover:- Why safety & certification are the hidden blockers for robotics at scale- How Saphira’s agent‑based platform turns CAD files, schematics, and risk regs into an actionable compliance plan- Real‑world wins (industrial arms, home humanoids) and what “shift‑left safety” looks like in practice- Career twists from Taekwondo black‑belt to hedge‑fund engineer to startup founderHit play and let me know what resonated most. 👇 All links & resources below.🔗 Guest linksWebsite – https://saphira.aiEmail – akshay@saphira.aiLinkedIn – https://linkedin.com/in/akshaychalanaX / Twitter – @Ac2zoom🎙 Connect with meLinkedIn – https://www.linkedin.com/in/greg-toroosian

We’ve officially hit Episode 100! 🎉 I sat down with Anthony Jules, co-founder & CEO of Robust AI, to unpack how Carter—the cart-shaped AMR that workers can literally steer—delivers a 60 % productivity boost for DHL without changing the warehouse layout.Talking pointsWhy “agency” (humans grabbing the robot’s handle) was the missing piece in previous AMRs.Vision-only navigation, no LiDAR: the leap to 4th-gen mobile robotics.Human-centric UX as the real adoption moat.Hiring for close-to-the-metal engineers and culture-first teams.What 10 years of collaborative automation might look like across logistics, healthcare, and retail.Links & resourcesRobust AI — https://www.robust.ai/Robust AI on LinkedIn — https://www.linkedin.com/company/robust-ai/DHL × Robust AI partnership (60 % productivity) — https://www.dhl.com/us-en/home/press/press-archive/2024/dhl-supply-chain-partners-with-robust-ai-to-drive-the-future-of-warehouse-automation.html“People + Physical AI” demo (YouTube) — https://www.youtube.com/watch?v=IFO_B43EDOcAnthony's LinkedIn: https://www.linkedin.com/in/anthonyjules/Connect with me — https://www.linkedin.com/in/gregtoroosian/Thanks for riding with us through a full century of episodes—here’s to the next hundred!

In this conversation I sit down with Zac Engler, founder of Bodhi AI, to talk about turning generative-AI hype into practical wins for smaller companies—and why “shadow-AI” bans often backfire. We cover:The moment Zac realised a recruiter could build a full-stack product in three weeks for < $1 k.Tactical first steps for any team: discovery sprint → 30/60/90 quick-win plan → agentic workflows.The trifurcation of work (AI-only, human-in-loop, human-only) and how to spot the right bucket.Why vertical, hyper-niche agents will beat one-size-fits-all models.What an 18-month adoption window really means—and how to avoid being “automated chaos.”Links:Bodhi AI – https://bodhiai.ioForbes feature: “AI’s Trifurcation of Work” – https://bit.ly/ZEForbes-AITrifurcationTurning On Machines (Zac’s forthcoming book) – https://www.zacengler.com/buy-bookMN AI Tools directory – https://mnaitools.comConnect with Zac: https://www.linkedin.com/in/zacengler/Connect with me: https://www.linkedin.com/in/gregtoroosian/

In this episode I sit down with Andy Williams, EVP Sales – North America at Exotec, to unpack how a single “mobile AS/RS” platform is helping retailers, 3PLs and grocers keep pace with labor shortages and same-day delivery promises. We cover:Why Exotec’s Skypod robots store, climb and travel like no other goods-to-person system.The shift from legacy conveyor projects to agile, software-defined warehouses.What “mobile AS/RS” really means for densification, throughput and expansion.Andy’s lessons on selling eight-figure automation to Fortune-500 operators.Where AI vision and data will push fulfilment over the next five years.Recent news • New Oxford Industries flagship installation → https://www.exotec.com/news/exotec-announces-partnership-with-oxford-industries-to-implement-next-generation-skypod-system-in-new-multi-brand-distribution-center/ • Next-gen Skypod launch → https://www.exotec.com/news/exotec-launches-next-generation-of-skypod-system-an-all-in-one-robot-based-as-rs-that-addresses-the-majority-of-processes-within-a-warehouse/Guest info Andy Williams – EVP Sales, North America, Exotec LinkedIn: https://www.linkedin.com/in/andrewtaylorwilliams/Connect with meLinkedIn: https://www.linkedin.com/in/gregtoroosian/

In this episode I sit down with James Emerick, co-founder & CEO of Cosmic Robotics, to hear how his family’s four-generation construction roots and a detour into autonomous heavy equipment sparked a new kind of job-site robot. Cosmic’s first product—the “Cosmic 1A”—lifts and places solar panels in 100-degree deserts and icy plains, giving short-staffed EPCs a safer, drop-in replacement for back-breaking manual installs. We talk through the realities of outdoor perception, why a “slightly smarter piece of equipment” beats factory-in-the-desert grand plans, and how mobile manipulation on Earth can pave the way for building off-planet.We covered:- James’s path from eighth-grade “future engineer” to early employee at Built Robotics- Three criteria Cosmic uses to spot high-leverage construction tasks- Designing for millimetre-level accuracy amid glare, dust and 50 °C heat- Digital-twin spin-offs and why customers ask for them sooner than you’d think- The labour crunch on solar farms—and the case for tools that amplify crews rather than replace them- Humanoids, hype cycles and what really matters in field roboticsFurther reading & resources:TechCrunch profile: https://techcrunch.com/2025/04/16/cosmic-robotics-is-building-robots-to-speed-solar-power-deployments-for-data-centers/The Robot Report funding story: https://www.therobotreport.com/cosmic-industries-gets-funding-to-automate-accelerate-solar-installation/Guest infoJames Emerick — Co-Founder & CEO, Cosmic RoboticsLinkedIn (personal): https://www.linkedin.com/in/jamesemerick/LinkedIn (company): https://www.linkedin.com/company/cosmic-robotics/?viewAsMember=trueWebsite: https://cosmicrobotics.com/Connect with me on LinkedIn: https://www.linkedin.com/in/gregtoroosian/

In this episode I sit down with Grace Kay Matelich, a philosopher-turned-operator who went from running Montana’s suicide-hotline to scaling AI data-ops at Invisible Technologies—and now to co-founding Bloomscroll, a “light-mirror” alternative to doom-scrolling. We unpack why recommendation engines are already an alignment problem, how subscription-based curation can flip the incentive model, and what it will take to put humans—not algorithms—back in the driver’s seat.We covered:- Lessons from crisis-line work, longevity research, and hyper-growth ops- Training large language models without cutting ethical corners- The birth of Bloomscroll and its “empowerment algorithm”- Why presence, autonomy, and intentional screen time beat infinite feeds- Practical tips for founders on hiring for grit and wielding AI as a toolGuest Info:Grace Kay Matelich — Co-Founder & CEO, BloomscrollWebsite: https://bloomscroll.com/Blog: https://bloomscroll.substack.com/LinkedIn: https://www.linkedin.com/in/grace-kay-matelich/Connect with me on LinkedIn: https://www.linkedin.com/in/gregtoroosian/

In this episode I sat down with serial entrepreneur Yegor Anchyshkin, co‑founder & CEO of Instock, to break down why small‑ and mid‑size fulfillment sites need a radically different ASRS architecture. Yegor traced his journey from founding Viewdle (AI video‑search; later acquired by Google) to launching Ukraine’s first online‑grocery service and leading Takeoff Technologies—experiences that shaped Instock’s “one lightweight robot, static grid” design. We dug into:- Why traditional ASRS can’t hit target ROI in micro‑fulfillment- Designing systems where operators hot‑swap a 90‑lb robot with no technicians- The hidden cost of on‑site debugging—and how simulation‑first culture fixes it- Hiring for grit and end‑user empathy in hard‑tech teamsWebsite: https://instock.com/YouTube Channel: https://www.youtube.com/@instockcomToronto 2024 install video: https://youtu.be/7tu5D1yNWEQIntro to Instock’s core ideas (2020): https://youtu.be/kXKRT9tBHMsLinkedIn: https://www.linkedin.com/in/anchyshkin/Connect with me on LinkedIn: https://www.linkedin.com/in/gregtoroosian/