Cloud Solutions × SonoSync — Samsung Medison
Real-time ultrasound streaming and remote machine control, built without ffmpeg.
Overview
My first project at Samsung, working with the R&D team on Samsung Medison’s Cloud Solutions — a collaboration with SonoSync to bring real-time ultrasound streaming to devices outside the scanning room.
The problem it solves is a practical one in clinical settings. An ultrasound examination happens at the machine, but the people who need to see it — a consulting specialist, a supervising physician, a remote colleague — are often somewhere else entirely. The system streams the live scan to a doctor’s desktop or other device as it happens, so distance stops being a constraint on who can observe an examination.
The scope went beyond viewing. We integrated the machine’s engines into the cloud, so medical staff can operate the ultrasound device’s components and use its features remotely, without being physically present in the same room as the hardware.
What I did
- Served as the streaming technology specialist on the team, responsible for publishing the generated ultrasound stream to the web in real time
- Built a custom application that removed the project’s dependency on ffmpeg for reading and displaying the stream
- Implemented double buffering using DirectShow and Media Foundation to control the frame pipeline directly
- Worked on the cloud integration that allows remote control of ultrasound machine components and features
- Contributed to the real-time transport layer connecting the machine to remote clients
The ffmpeg problem
The original pipeline relied on ffmpeg to read and display the generated stream. In a clinical context, that dependency carried real costs — added latency, licensing complications, and limited control over buffering behaviour at exactly the point where dropped frames matter most.
Replacing it meant building an application that reads and renders the stream directly, with double buffering implemented through DirectShow and Media Foundation. Removing ffmpeg from the pipeline eliminated the licensing and latency overhead, and put frame handling under our own control rather than behind a third-party abstraction.
Skills
| Technical | Windows application development, computer networking, real-time transport protocols, C++, C#, Python, JavaScript, machine learning, Docker, cloud solutions using AWS and Azure, FFmpeg, MediaMTX, WebRTC, libdatachannel, named pipes |
|---|---|
| Soft | Communication, negotiation, time management, task and responsibility management, teamwork, collaboration, continuous self-improvement |