Michal Sofka
Building AI software systems at startups


I am passionate about building software, specifically around machine learning, computer vision, and medical imaging. Since 2017, I have been making Hyperfine's Swoop portable MRI smarter, and I am currently director of the AI and cloud teams. Previously at Cisco, Siemens, and RPI, see my journey below.

In each of my previous roles, I started something out of nothing (in addition to my main job) which grew and ended up in a product. I like taking calculated risks and taking on life adventures. I drove from New York to Cancun once.


« more »

Recent Tweets

« more »


Hyperfine developed highly portable MRI system that wheels directly to the patient’s bedside, plugs into a standard electrical wall outlet, and is controlled via a wireless tablet such as an Apple iPad. I started building out the machine learning competency when we were 6 employees and had first imagages from the scanner. I now lead the deep learning team to advance image reconstruction and interpretation of MR images. You know you are lucky when you work with incredibly talented team that always remembers to have fun. Recent result from our lab: Deep learning reconstruction of a T2 scan from a portable MRI system and a conventional 1.5T scan. Absence of shielded room, lower field strength, and simpler electronics are compensated for by a smart computational imaging pipeline.

Butterfly Network

Butterfly built a handheld ultrasound device that plugs into an iPhone, costs under $2000, and sends data for review and reporting. Collecting data with the first handful of prototype probes, our team worked on tools for image acquisition assistance and image interpretation. It is inspiring to see the team relying on modern software tech stack and always pusing the boundaries of what's possible in all key areas of the company.


Cisco acquired university startup Cognitive Security that built machine learning tools for network threat defense that analyzed billions of network traffic logs every day. I lead the the design and development of new machine learning algorithms for detecting and classifying new types of malware and for building label database from handful of manually analyzed threats. It all started when a tight group of smart scientists and engineers got together and started building new engines that are now critical for Cisco's security business and a number of products.


Siemens innovation center in Princeton is a home to scientists and engineers focused on researching and developing emerging technologies in healthcare to strengthen the company portfolio of medical imaging and laboratory diagnostics, therapeutic and molecular diagnostics, managed services, consulting, and healthcare IT services. I joined a team that was pioneering machine learning tools for diagnosis and treatment planning. I lead and contributed to many products, including automatic measurements in fetal ultrasound, detecting and finding outlines of anatomical structures in CT scans, and building software tools for total knee replacement surgery (best product award from J&J).

Rensselaer Polytechnic Institutue

I got a PhD and MSc in Computer Science from Rensselaer Polytechnic Institute (RPI). My thesis was on retinal vessel extraction, uncertainty in point and image registration, multimodal and illumination invariant keypoint descriptors, and location registration and recognition for mapping regions in CT scans.