DermaSense partnered with Shearwater Data to pioneer an in-pharmacy smart kiosk capable of rapid dermatological triage. Leveraging embedded multi-spectral cameras and quantized convolutional neural networks running on localized edge GPUs, the system classifies 14 distinct epidermal conditions, checks for suspicious melanocytic lesions, and formulates clinician-validated skincare treatment protocols without transferring unencrypted biometric imagery over public networks.
Industry
Digital Health, Computer Vision & Clinical Diagnostics
Long Waitlists and Inconsistent Triage in Primary Care
With outpatient dermatology appointment wait times averaging over 38 days across metropolitan areas, early-stage skin conditions and malignant melanomas were routinely diagnosed far too late.
DermaSense required an autonomous, highly reliable clinical scanning device that could be placed in neighborhood pharmacies, capable of operating with zero perceptible delay and absolute patient data isolation.
ENGINEERING INTERVENTION
Quantized TensorRT Models & Local Hardware Enclaves
Shearwater Data built a hybrid dual-backbone vision pipeline. High-resolution raw frames are captured under polarized ring lighting and fed directly into an on-device NVIDIA Jetson Orin compute unit.
Using FP16 and INT8 quantized vision transformers, the kiosk achieves 99.1% agreement with board-certified dermatologists on skin texture, pore dilation, erythema, and dyspigmentation within 42 milliseconds. Sensitive patient imagery never exits local RAM; only anonymized vector telemetry is synchronized to the central lakehouse.
“The engineering rigor Shearwater Data brought to our edge AI kiosks transformed an experimental clinical vision model into an industrial-grade medical diagnostic tool deployed across 180 pharmacy locations.”
Dr. Elena RostovaChief Medical Officer, DermaSense Health
MEASURABLE IMPACT
Empirical Results
Across initial retail pharmacy rollouts, the kiosks delivered decisive clinical and commercial outcomes: