Hardware Podcasts

83. Ultrasound Scanning Stickers & Autonomous Protein Discovery

The Next Byte Podcast · 16 Aug 2022 · 32m

AI in practice: one real use of AI described

What this episode covers

Hosts Daniel and Forbode discuss an MIT innovation: a postage-stamp-sized ultrasound sticker featuring rigid transducers embedded in elastomer layers to maintain image quality and hydration during long-term wear. They then examine a Princeton and Rutgers collaboration that uses machine learning and Hamilton M Star liquid handling robots to autonomously discover stable protein interactions, accelerating drug and material development. The hosts highlight how this automation solves trial-and-error bottlenecks in chemistry, enabling the discovery of molecules far more effective than human-designed ones.

AI in practice

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

Engineering & simulationExpanded catalog to 500,000 molecules

Machine learning was used to identify viable chemical candidates by training on 1,500 combinations of 500 molecules and 3 proteins, allowing the system to predict interactions for 500,000 molecules without explicit training on the larger set.

“The 1st ingredient was this machine learning, the other way that they're helping attack this challenge is actually physical automation. So the 1st part was kind of intellectual automation, right? Teaching a machine to learn and to think the way that we would do to try and solve the problem.”

Speaker 2 · Hear it at 20:22

Most relevant to

early-stage companies

Topics

Mentioned

Companies
MIT
Tools
Hamilton M Star

Related episodes from other shows

Other shows on the same ground, closest first.

  • Catalyst · 36m

    Inside a $300 million bet on AI for physical R&D

    Ekin Dogus Cubuk, co-founder of Periodic Labs, explains how his company combines frontier AI models with automated physical labs to accelerate materials discovery. Cubuk argues that recent advances in reasoning models like OpenAI's O1 and improved lab automation allow AI to overcome its reliance on limited training data by generating synthetic data and running high-throughput experiments. The company is targeting breakthrough materials, specifically high-temperature superconductors, using a hybrid approach where humans provide initial hypotheses and AI handles execution and iteration. With a $300 million seed round, Periodic Labs aims to build a generalized scientific intelligence platform that serves both its own discovery goals and external partners' R&D needs.

    AI & machine learningMaterials & manufacturing processes
    10 chapters

  • Hackaday Podcast · 1h 11m

    Ep 314: It's Pi, but Also PCBs in Living Color and Ultrasonic Everything

    Hosts Elliot Williams and Dan Maloney review a week of hacks, starting with a sound-guessing game won by 'angle grinder'. They discuss Alex Toussaint’s Supercon talk on a 380-microphone sonar array capable of identifying mosquitoes via Doppler shift, and Joe Long’s exploration of full-color PCB silkscreens from JLCPCB and PCBWay for badge art and assembly guides. The episode also covers an automated machine for embedding fiber optics in concrete, a monolithic BMS for prismatic battery packs, and ultrasonic knives for cutting food.

    Sensors & measurement in productsEDA & CAD tools
    7 chapters

  • Robot Talk · 32m

    Episode Eight - Soft Robotics: a new kind of robot

    Dr. Thomas Thuruthel of the University of Cambridge explains how machine learning models control soft robotic grippers and describes self-healing elastomers with dynamic cross-links for disaster and medical applications. Angus Clark from Imperial College London discusses malleable robotics, using variable stiffness links like Mylar to combine the strength of rigid robots with the compliance of soft ones for industrial collaboration. Dr. Jelizaveta Konstantinova of Ocado Technology details how soft robotics and AI-driven suction systems handle diverse grocery items, leveraging acquisitions of Kindred Systems and Paddington Dynamics to scale automation.

    Robotics & autonomyMaterials & manufacturing processes
    5 chapters

  • Hackaday Podcast · 48m

    Ep 126: Cable 3D-Scanner, Tesla Charger Robot, Ultrasonic Anemometer, and a Zoetrope

    Hackaday editors Mike Szczys and Elliot Williams review a week of hacks, focusing on Leonard Wouters' reverse engineering of the Starlink terminal firmware, which revealed geofencing and hardware locks. They discuss Harrison McIntyre's 3D zoetrope that uses overlapping frames to double animation resolution, and Scott Rumschlag's submillimeter mechanical 3D scanner that uses a spooled wire and gimbal for high-precision measurement. The episode also covers Pat Larson's garage robot that autonomously plugs in a Tesla, an ultrasonic anemometer project, and news regarding China's Zhurong Mars rover.

    Debugging & failure analysisPrototyping & iteration
    8 chapters

  • Robot Talk · 16m

    Episode Twenty-Nine: Sina Sareh

    Sina Sareh, Academic Leader in Robotics at the Royal College of Art (RCA), discusses his lab's focus on human-centric robotics, including autonomous systems for offshore wind turbine inspection and adaptive grippers inspired by animal attachment mechanisms. He details the MINRI consortium project involving a crawler, UAV, and autonomous boat, as well as his earlier work at Imperial College London on soft robotic manipulators for surgery and additive manufacturing. Sareh explains how deep learning enables real-time decision-making for anchoring modules to handle unpredictable natural and recycling environments.

    Robotics & autonomySensors & measurement in products
    5 chapters

The summary, chapters and topics are generated from the episode’s audio by this directory, not written by The Next Byte Podcast. The episode itself belongs to its publisher.