Annie Katz

“An Apple Watch for the gut.”

An overnight wearable: three piezoelectric contact microphones and a thermistor sewn into an abdominal wrap, run by a QT Py ESP32‑S3 - that monitors gut motility and circadian rhythm noninvasively, at home. Clinical tests cost thousands and give a single snapshot; my device is reusable at ~$50 a unit.

I designed the full analog signal chain (AC‑biased mics → buffers → amplification → separate ADC channels), learned KiCad and laid out the two‑layer SMD board myself, and shipped production Gerber/BOM/CPL files to JLCPCB. Python analysis (NumPy · pandas · SciPy) used Welch power spectral density to resolve gut slow‑wave rhythms at 0.01-0.05 Hz against sleep stages. Now in human testing with GI clinics.

An at‑home ultrasound wand I built after re‑tearing my hamstring.

I streamlined a commercial 128‑transducer probe down to 8 piezoelectric transducers I cut and sanded by hand, driven by a 5 MHz PWM signal. I wrote C firmware (with DMA) to fire pulses and record echoes, then a Python pipeline that composites three captures to raise signal‑to‑noise and reconstruct the image.

To train it, I cold‑emailed 200+ radiologists and sports‑medicine doctors worldwide to source 800+ ultrasound images (no central hamstring database exists), and worked with two radiologists to confirm tear grades. Two CNNs, lesion grade then specific hamstring muscle, served in a website I built.

A two‑year project for my sister, who has dyspraxia.

Year one, Grip Hero: a motorized assistive glove (H‑bridge, Arduino Nano, Kevlar tendon cords, 3D‑printed gearbox) that raised her measured grip from 13.4 to 20 lbs, the level OTs consider functional for everyday tasks.

Year two, Gripster: seeing it help but not fix the underlying weakness, I pivoted to a resistance rehab trainer that builds strength permanently: a waterjet‑cut aluminum frame I machined on a lathe, a load cell with PID‑controlled motor resistance, and an ESP32 acting as a Bluetooth keyboard so squeezing the grip plays a browser game I wrote, with an Epley‑formula protocol, Chart.js progress tracking. Multiple physical therapists want to trial it.

A reusable post‑surgical infection monitor, inspired by my grandmother.

A machine‑washable elastic wrap with a modular electronics layer and four hand‑soldered thermistors. It compares average peri‑wound temperature to surrounding skin and flags a >2 °F differential as an early infection sign through a web interface. Originated for my grandmother's post‑mastectomy infection, and it won the Grand Prize out of 18 teams at Yale.

A smart cast that makes what's happening under the plaster visible.

A foot‑and‑ankle surgeon told me casts are almost never opened until something is already severe, an infection or ulcer, and by then the damage is done. So I built force‑sensitive resistors and a thermistor into the cast lining at pressure hotspots (calcaneus, malleoli) to catch pressure buildup and temperature‑based infection signs early, and alert the patient, caregiver, and doctor. A magnifying glass into the cast.

A wearable ML misophonia aid for my brother.

A Raspberry Pi Compute Module 4 runs a TensorFlow Lite audio classifier that detects trigger sounds in real time and shows a Positive Attributable Visual Source on a small display, reframing the sound to blunt the fight‑or‑flight response, based on published PAVS research. Developed with mentorship from the Duke Center for Misophonia and Emotion Regulation, which selected me into its PRIMER workshop and invited me into its weekly research meetings.

A stylish beanie that delivers discreet headache therapy.

Discreet heat and vibration therapy for headaches, controlled by capacitive touch sensors sewn from conductive thread, or a web app, and run by an ESP32. I led the team, built all the electronics and fabrication, and taught my teammates to work with e‑textiles.