Can you discuss your background and career journey?

My journey has been a bit unconventional, shaped by both medicine and data science. I originally trained as a medical doctor, but along the way I developed an intense curiosity for the data behind clinical decisions. During medical school, I found myself drawn to research and technology for managing data and even pursued a degree in mathematics on the side. This wasn’t part of some grand plan; it happened organically as I kept encountering medical problems that begged for data-driven solutions. I often say I became a “mathematician by accident” because my adventure into math was a byproduct of trying to analyze patterns in clinical data.

Since you mention the phrase “a mathematician by accident,” can you share how you came to realize the power of data in healthcare, and how did it change your professional path?

I never set out to formally become a mathematician – if you told 20-year-old me that I’d be deep in algorithms and machine learning models, I’d have been surprised. My plan was to be a practicing physician. Realizing the power of data transformed my career trajectory. Instead of following a traditional residency path, I decided to focus on medical data science. I dove into projects where I could merge clinical insight with data analysis. This shift ultimately changed my professional identity – from being a physician who dabbles in data, I became a data scientist who never forgets the human context behind the numbers. It led directly to co-founding IOMED.

As co-founder, can you discuss what IOMED is and how you hope it can impact healthcare research?

I wanted to take that idea – that data can save lives and improve healthcare – and build a platform to make it a reality. In essence, I went from prescribing medications to prescribing data-driven decision making. It was a leap of faith at the time, but I have no doubt it was the right move. At IOMED, we partner with hospitals to maintain their OMOPs up to date and we enrich them with data from non-structure sources like clinical notes and discharge reports using AI. The healthcare industry generates an immense amount of data, but much of it sits unstructured and siloed. Our goal at IOMED is to unlock that data and make it useful while of course respecting patient privacy. To do so, we partner with the industry to execute studies in an efficient and timely manner, offering a solution that makes requesting data, getting ethics committees approval, quality assurance and delivery processes as fast as possible. By lowering the barriers to access and analyze health data, we aim to enable discoveries that improve patient care. In the long run, we believe this kind of platform can support everything from public health surveillance to clinical trial optimization. It’s about materializing the technology for making the European Health Data Space a reality.

Your team served as an SME in the EHDEN project. What were the biggest challenges to gaining expertise with data mapping, and how has that expertise impacted your company?

Engaging in the EHDEN project was both exciting and challenging for us. One of the biggest challenges we faced was dealing with the incredible diversity of healthcare data. Every hospital and health system has its own way of recording information. Different electronic record systems, different coding schemes, even different medical jargon in clinical notes. To convert all that into a common format (OMOP), you need deep knowledge of both the source data and the target standards. We had to become fluent in various medical terminologies (ICD, SNOMED, LOINC, etc.) and understand the nuances of each dataset. It was a steep learning curve. Going through the EHDEN certification and projects essentially turbocharged our capabilities. We emerged with a much stronger command of the OMOP vocabularies and the best practices for mapping. As a result, our internal tooling for bringing new hospitals’ data onboard became far more efficient and reliable. We’re almost fanatical about data quality now (in a good way!). In short, the challenges of EHDEN made us stronger. They pushed us to refine our technology and practices, and that expertise continues to propel our company forward in enabling trustworthy, large-scale health data analysis.

IOMED highlights numerous use cases that can benefit from the data network you have assembled. To impact such a wide breadth of work (population profiling, observational studies, patient journey, etc.), how critical are advanced analytics to get the maximum potential from this data?

They are absolutely critical. Having a large, well-curated dataset is the foundation; we often say that data is only as valuable as the questions you can answer with it. Each of those use cases demand sophisticated analytical approaches and quality assurance processes to truly extract value. We help data users execute those use cases together with the data holders in the network while providing support on all the data quality assurance, checks with the hospitals on the data, and, more importantly, on the nuances of the clinical practice (and recording of data) of each of the hospitals.

Your poster at the 2024 Europe Symposium highlighted NLP-derived OMOP results. Can you discuss the value of and opportunities provided by NLP in observational research?

Natural Language Processing (NLP) opens up a world of opportunities by unlocking unstructured clinical data, which is a massive portion of what’s recorded in healthcare. It’s often cited that around 80% of healthcare data is unstructured – clinical notes, hospital discharge summaries, pathology and radiology reports, etc. Traditionally, observational research has relied mostly on structured data (diagnosis codes, medication lists, lab results) because those are easier to tabulate and map. But think about all the rich detail in a physician’s narrative: descriptions of symptoms, lifestyle factors, assessment of how a patient is doing beyond the numbers. NLP allows us to systematically incorporate that wealth of information into research datasets. At IOMED, we use NLP to derive structured data from text and integrate it into the OMOP common data model. The value of this is huge. For one, it improves phenotyping – identifying cohorts for studies with specific criteria. It’s especially useful for outcomes or exposures that aren’t well-captured by billing codes. For example, consider symptoms or procedures that are often recorded in these notes (“patient experienced had a NYHA score class III with dyspnea…”), but there may not record the specific code for that. NLP can pull those observations out, giving researchers a much fuller picture of a drug’s real-world effects. Our poster at the OHDSI Europe Symposium 2024 presented a framework for verifying and validating data captured via AI (our NLP in this case) in the context of OMOP. This is an important point – as we use NLP to generate research data, we need to ensure its reliability. We showcased how we cross-verify NLP-extracted data points against known structured data and clinician review, to quantify accuracy. The results were very encouraging, indicating that with proper validation, NLP-derived data can be trusted alongside manually curated data.

What are some of your hobbies, and what is one interesting thing that most community members might not know about you?

In my free time, surfing gives me a sense of flow and presence that clears my head and reconnects me with the natural world, while meditation helps me center myself. An interesting fact about me is that at 17, I co-founded an art gallery in my home town, a space dedicated to urban art. We supported local graffiti artists and skaters, organized public exhibitions, and fostered a creative community around street culture. It was my first experience building something from the ground up with cultural and social impact.