Hardware Podcasts

Live Recording: the Missing Data for Physical AI

Go/No-Go · 8 Oct 2026 · 45m

AI in practice: 3 real uses of AI described

What this episode covers

Eduardo Torrealba of Lumafield, Vikram Pavate of Tacta Systems, and Pari Singh of Flow Engineering discuss how physical AI is bottlenecked by the lack of specific, non-public data. They argue that while software AI thrives on open internet data, physical products require proprietary datasets from CT scans, tactile sensor gloves, and engineering context to train effective models. The panel highlights the 'missing middle' of manufacturing—products costing $5 to $500—as the key area where automation and AI can enable reshoring to the US. Each founder shares a 'punched in the mouth' moment: Lumafield's scanner durability issues in unconditioned factories, Tacta's glove being rejected for being too soft, and Flow Engineering's decision to burn eight years of code to rebuild for an agentic future.

AI in practice

Where this conversation describes AI being used to build or deliver a product. Part of the AI in practice tracker.

Engineering & simulationFlow Engineering

AI agents automatically verify hardware designs by detecting how changes in CAD, simulation, or firmware break specific engineering requirements, aiming to reduce development time and cost.

“our AI systems engineering agents can listen to change in CAD, in simulation, and get in firmware, and then automatically realize that a change in one place affects a change in another, which effects a change in another, that breaks a requirement.”

Pari Singh · Hear it at 3:05

Engineering & simulationTacta Systems (Dexter AI model)

An AI model is trained on tactile and motion data captured from workers wearing sensor-equipped gloves to generate control policies for dexterous robotic hands.

“we take that data to train our AI model, we call it Dexter AI model, generate policies that then go back and use they're not deployed on those robotic canes to do these dexterous tasks.”

Vikram Pavate · Hear it at 4:31

Engineering & simulationFlow Engineering30 to 106 customers in 6 months

The company rebuilt its entire codebase and product after realizing that AI-native development practices were necessary to compete with rapid hardware engineering cycles, leading to exponential growth in customers and revenue.

“we spent 6 months to rebuild the entire company. So 100% the code went on the bin. We started a code base again. … we've gone from 30 customers to 106 months. Like, it's just explode, like revenues exploded.”

Pari Singh · Hear it at 19:59

Most relevant to

early-stage companies

Topics

Mentioned

Companies
Lumafield, Tacta Systems, Flow Engineering, Matter Venture Partners, Oso Labs, Form Labs, Lux View, Lux Systems, BAE, The Engineering Company, Rivian, Sequoia, SpaceX, BYD, Tesla
Products
Neptune, Triton, Mars, Falcon 9, Starship, Giga Factory
Also
Jensen

Related episodes from other shows

Other shows on the same ground, closest first.

  • The Manufacturing Executive · 25m

    AI in Manufacturing: Fusing Human Knowledge and IIoT Data w/ Christopher Nguyen

    Christopher Nguyen, CEO & Co-Founder of Aitomatic, explains why effective AI in manufacturing requires fusing IIoT data with the tacit knowledge of frontline workers. He argues that pure data-driven models fail in physical domains due to the 'data wall' and lack of failure examples, necessitating the capture of domain expertise. Nguyen discusses how his software helps industrial companies translate unstructured human knowledge into structured forms to improve defect detection and process optimization.

    AI & machine learningConnected products & IoT
    7 chapters

  • The Manufacturing Executive · 32m

    The Hard Truths About Reshoring American Manufacturing w/ Eddy Azad

    Eddy Azad, CEO of Parsec Automation, discusses the strategic realities of reshoring manufacturing to the United States, arguing that it is a long-term, complex process rather than a simple cost-saving switch. He explains that reshoring makes the most sense for high-value, high-tech products like semiconductors and medical devices where supply chain resilience and innovation are critical, rather than for commoditized goods. Azad highlights that while automation and AI can mitigate higher US labor costs, they require robust data infrastructure and end-to-end integration to be effective. He also addresses the impact of AI on the workforce, noting a shift from manual execution to human-in-the-loop oversight and prompt engineering.

    Policy, trade & tariffsAI & machine learning
    8 chapters

  • The Manufacturing Executive · 27m

    Moving From Generative to Physical AI on the Manufacturing Floor w/ Etienne Lacroix

    Etienne Lacroix, Founder and CEO of Vention, discusses the transition from generative AI to 'physical AI' on the manufacturing floor, defining it as injecting AI code into embodied systems like robots to achieve autonomy. He identifies bin picking and vision-guided robotics as the only two task categories currently ready for industrial deployment, citing specific metrics like 99.9% first-pick success rates and sub-two-year payback periods for two-shift operations. Lacroix contrasts these mature applications with the current limitations of Vision-Language-Action (VLA) models, which require excessive post-training hours, and highlights the operational challenges of non-deterministic robot movements.

    AI & machine learningRobotics & autonomy
    6 chapters

  • The Robot Report Podcast · 1h 19m

    Inside Singapore's Physical AI Revolution

    Lionel Lim of the Singapore Economic Development Board, Raahul Kumar of Certis Group, and Zhihao Chin of Home Team Science & Tech Agency discuss Singapore's strategic push into physical AI, highlighting its role as a global robotics hub with over 300 companies. They detail the ecosystem's strengths, including a strong manufacturing base, government incentives, and testbeds like the Jurong Innovation District and Punggol Digital District, which support companies from prototyping to scaling. The guests emphasize the importance of integration, reliability, and scalability in deploying robots in unstructured environments, citing deployments by Certis and HTX in security and public safety. They also address the challenges of data collection, talent development, and the need for clear use cases to move physical AI from experiments to mainstream adoption.

    Scaling productionRobotics & autonomy
    7 chapters

  • The Robot Report Podcast · 51m

    Building robots that survive the warehouse

    Josh Cloer, General Manager for North America at No Magic, explains how the company applies 'physical AI' and vision-language-action (VLA) models to warehouse each-picking and recovery workflows. He emphasizes that No Magic’s competitive advantage lies in using real-world production data rather than simulation to handle edge cases, such as flexible shoebox handling. The discussion covers the shift from batch to continuous 'always-on' operations, the integration of AI layers onto existing hardware like ASRS, and strategies for de-risking automation adoption for supply chain leaders.

    Robotics & autonomyAI & machine learning
    9 chapters

The summary, chapters and topics are generated from the episode’s audio by this directory, not written by Go/No-Go. The episode itself belongs to its publisher.