Where this conversation describes AI being used to build or deliver a product. Part of the AI in practice tracker.
Engineering & simulationTexas Instruments MSP430 (microcontroller)5.4 millijoules per process; 1.3 millowatt draw
The team selected an off-the-shelf microcontroller and optimized the sampling rate and model architecture to enable on-device machine learning inference with extremely low power consumption.
“I'm actually pretty certain that the microcontroller they use. It's like a Texas Instruments MSP 4 something, it's actually available on Mauser. … they change the sampling rate of the audio to match that of the analog to digital converter on the microcontroller that they wanted to use.”
Engineering & simulationAutonom US handheld surgical robotUsers with no experience successfully inserted needles in under one minute
The device uses AI to process noisy ultrasound data and simplify it into a visual interface (rectangle and crosshairs), allowing untrained users to locate blood vessels and insert needles with expert-level success rates.
“It uses ultrasound to help locate the blood vessel, and it uses AI to simplify that data from the ultrasound into a display that is, you know, makes sense to a lay person. … it takes this complex, noisy data and turns it into something really simple…”
Engineering & simulationThomas Neisler's speech recognition algorithmProvided 100s of early warnings; deployment in under 10 minutes
A team used a custom speech recognition algorithm to listen to rural radio broadcasts, identify trigger words for emergencies, and categorize events to trigger automated alerts to city centers.
“using his speech recognition algorithm from his past, it looks for specific trigger words like malaria or flood or other, you know, natural disasters or diseases. It categorizes events as either this is an emergency, you need to respond now…”