The Distillery
The projects described in this portfolio are primarily closed source, unpublished work but are current areas of focus and in some cases side learning. These are engineering projects across mobile, embedded GPU, and clinical informatics domains on macOS · iOS · watchOS · Ubuntu · NVIDIA Jetson Orin Nano.
Implementation of a B-mode ultrasound pipeline brought up on NVIDIA Jetson Orin Nano hardware. Includes live display to viewers from shell driven simulation based on ROS 2, and Holoscan substrates — with native iOS & macOS DDS viewport applications.
Realization of a 3D volume rendering engine for macOS and iOS. Implemented using Metal shaders and includes GPU ray marching through MTLTexture3D with gradient lighting, a configurable transfer function pipeline. Includes JPEG, PNG as well as NIfTI ingestion (DICOM slice on deck).
This is a dry run at a Retrieval-Augmented Generation vertical — fetches and embeds NIH UMLS vocabularies and DICOM DCMR code concepts into PostgreSQL + pgvector based corpus. Can generate clinical seed reports that (optionally) can contain image references, aligned with HL7 FHIR Diagnostic Module on iOS.
Functonal concept proof time series solution for Shake or Tremor Sensing based on detected motion samples from Apple Devices - iPhone and the Watch wearable. Ingredients include capure of Inertial Measurement Unit (IMU) samples for acceleration, rotation and gravity - passed to an inference pipeline, a lightweight 1D CNN. Clinical target: Parkinson's Disease and Essential Tremor detection.