Machine Medicine Is Building an Operating System for Neurotechnology
- Dominic Borkelmans
- 1 day ago
- 6 min read
In the last few years, the neurotech revolution, with at its center brain-computer interfaces, has been framed as a silver bullet for neurology. Yet in the clinic, the reality looks very different. Even traditional deep-brain stimulation (DBS) implants remain rare in patients, and new devices emerge carefully and slowly. Nowhere is this clearer than in movement disorders, where clinicians depend on subjective scales and in-clinic snapshots that struggle to capture the complexity of motor symptoms.
Jonathan O’Keeffe, founder and CEO of Machine Medicine, believes that the gap in how we measure and understand neurological conditions comes down to information. Most treatments are discovered by chance. But if the brain is an information-processing system, then progress depends on measuring, organizing, and acting on that information far more precisely than we do today. Machine Medicine’s core product, Kelvin, is built around that premise: rather than focusing on a single device or biomarker, it serves as an operating system for building the AI needed to unleash the field of neurotechnology in full.
From Philosophy to Neurology
While O’Keeffe is now committed to improving the field of neurology, he did not begin his career in medicine. His first degree was in philosophy, where he encountered mathematical logic, such as Gödel’s incompleteness theorem, and early ideas about computation and mind. Those questions led him to a broader one that still drives his work today: if our brains are computational systems, what kind are they, and what does that mean for what we are as humans, and what we potentially could be in the future?
“I didn’t really want to be an academic philosopher,” he says. “The output, papers and books, is interesting, but it didn’t feel like the place where the action was going to happen. As soon as philosophy makes progress, it becomes a science.” Artificial intelligence is one example where O'Keeffe’s instinct has proven correct: many of its foundations were laid by mathematicians and philosophers before it became a mature scientific field and eventually a sprawling industry.
To get closer to the action, he pivoted into medicine. He studied neuroscience as a medical student, completed clinical training, and worked as an NHS doctor for several years. Along the way, he added a Master’s in machine learning and a PhD in computational neuroscience, deepening the link between his philosophical interests and translational brain science.
Returning to neurology after his PhD, he realized the traditional clinical route was not the right fit. “I wanted to work on neurology, not in neurology,” he says. While many of his peers from computer science were starting companies, few doctors at that time were encouraged to take that path. O’Keeffe gave up his neurology training post anyway and launched Machine Medicine, convinced that better tools for measuring and applying brain-related data could ultimately have much more impact on patients than a single clinical career.
Information as a Therapeutic Medium
Machine Medicine builds tools that make neurological data easier to capture, interpret, act on, and ultimately build on. At its core is Kelvin, a platform that helps clinicians and researchers record high-quality data, especially video, manage it securely, and analyse it with new or existing machine-learning models. O’Keeffe wants to give teams the infrastructure to measure neurological and psychiatric symptoms and treatment effects with far more precision than subjective clinical scales allow.
Fundamentally, Kelvin functions as an operating system for building and deploying neuro-AI. Clinicians and researchers can capture data through mobile apps used in the clinic or at home, and upload other modalities such as MRI or device data through a web portal. API integrations pull information directly from neurostimulation devices and external systems. On top of this sits Kelvin Cortex, where teams can build ML models using HIPAA-secure patient data or deploy models that Machine Medicine has already trained. These include pose estimation tools, activity recognition models, and clinical evaluation models that generate structured assessments from recordings.
The platform is used by clinicians through a freemium model that allows them to contribute anonymized data in exchange for the software, or pay a small fee when data-sharing is not possible. Enterprise customers include pharmaceutical and medical device companies, from early-stage teams to large firms such as Sanofi developing new drugs. Machine Medicine also supports institutions working on ambitious neuromodulation projects, including groups developing stimulation systems for neglected neurological diseases.
Machine Medicine’s practical foundation reflects O’Keeffe’s view that the brain is, at its core, an information processing system, one whose failures we still struggle to measure clearly. For decades, many neurological and psychiatric treatments were discovered almost by accident, long before anyone understood how or why the underlying information flows were breaking down. If the field is going to mature, he argues, it needs tools that can actually capture and quantify those processes with the same rigor seen in other parts of biology.
Neuromodulation is one example of this shift away from exogenous approaches, such as drugs, and towards information streams. Devices such as deep brain stimulators interact with the nervous system using part of its own language, electrical signals, rather than introducing broad-acting chemicals.
"We're really trying to accelerate brain science and help shift it from a serendipity-driven field to one that is much more mechanistic and information-based."
“SSRIs aren’t something your body produces naturally. So rather than putting in foreign substances, we’re speaking to nervous systems in their own language,” O’Keeffe says. “We don’t yet understand the neural code, so in a sense we’re jabbering at it rather than speaking fluently, but the potential for real progress lies there.”
This logic extends beyond purely neurological and psychiatric conditions. O’Keeffe points to work like splenic nerve stimulation for rheumatoid arthritis, where modulating a peripheral nerve can influence immune pathways without the side effects of current drugs. It is not a neurological condition in the traditional sense, but the mechanism is informational: adjust the signals, and the immune system responds differently. For O’Keeffe, that is the direction of travel, treating disease by steering biological information flows, not just by suppressing symptoms.
The Machine Medicine Mission
Despite the philosophical roots of his thinking, O’Keeffe’s goals for Machine Medicine are pragmatic. Over the next decade, he wants the company to demonstrate clear, bottom-line returns from this data-driven approach. “Right now, many of our customers work with us as an act of faith,” he says. “They believe this approach will allow them to do things they otherwise couldn’t, but it’s still speculative. I don’t think many companies in this space have proven it yet.”
If platforms like Kelvin can show that better, richer data leads to more efficient trials, improved responder selection, faster regulatory paths, or better long-term outcomes, then adoption could shift from being vision-driven to market-driven. At that point, it would be difficult for companies not to use such tools.
Recently, there has been an emphasis on passive health monitoring. O’Keeffe sees value in these signals, but is skeptical that they can replace structured, active assessments. “You can see where the idea comes from, it’s an engineering mindset,” he says. “If you’re monitoring a bridge, you attach accelerometers and track movement. But for something as complex as Parkinson’s, that’s just a tiny aperture on the full picture.” Machine Medicine ingests passive data, including via platforms like Apple HealthKit, but O’Keeffe believes it complements, not replaces, modalities such as video and speech that clinicians can immediately understand and already use.
Initially, Machine Medicine focused where neurotechnology adoption is most advanced today: movement disorders, and Parkinson’s disease in particular. Since then, they have expanded into other conditions with motor components, such as dystonia, tardive dyskinesia, essential tremor, and multiple sclerosis, and ultimately, they will address disorders that are not classically neurological but are amenable to modulation through the nervous system.
Through all of this runs the same philosophy: there are no sudden revolutions in medicine, but there can be a consistent evolution. By treating biology as an information-processing domain and building the tools to measure, manage, and act on that information, Machine Medicine hopes to nudge neurology into a more mechanistic, data-literate future.
About the Founder
Jonathan O’Keeffe, MD, PhD, is the founder and CEO of Machine Medicine, where he leads the development of Kelvin, a data operating system for clinical-grade neuro AI. Originally trained in philosophy, he went on to study medicine and neuroscience, working as an NHS doctor before completing a PhD in computational neuroscience and a master’s in machine learning. His focus now is on building the software and data layer for bioelectric medicine, translating academic advances in computer vision, neurology, and neuromodulation into practical tools that make motor symptoms measurable in real-world clinical settings
About the Firm
Founded in London in 2014, Machine Medicine develops Kelvin, a neurodata platform that uses video, audio, wearables, active tasks, and computer vision to provide precise, scalable assessments. Kelvin allows users to capture multi-modal assessments, in clinic or at home, along with device or other external data, and then build and deploy machine-learning models on the combined datasets. The company positions itself as the data and software layer for next-generation brain-tech, working with health systems, pharma, and device makers to bring more rigorous, information-based measurement and modelling into neurotechnology.
Visit: machinemedicine.com