Concept electronic medical technology on tablet Digital healthcare body system analysis and networking on a holographic interface. science and innovation medical technology and network concept

[Adobe Stock]

The complexity of patient records, with their mix of unstructured notes and diverse data types, defies traditional analysis. While a thorough review of even a single patient’s file can be tedious  for a human, AI tools offer the power to analyze tens or even hundreds of millions of records, unlocking data patterns that would otherwise remain hidden.

A burgeoning number of companies are popping up to address the tasks of amalgamating and deciphering patient records using AI tools. Among them is nference, which has attracted backing of notable institutions such as the Mayo Clinic, Duke Health, Emory Healthcare, Vanderbilt University Medical Center and The Bill & Melinda Gates Foundation. In 2020, the company was highlighted as one of the top digital startups in the world.

Nference’s AI platform has a vast array of data at its disposal. In all, it captures more 11 million patient lives spanning more than 20 years. This data encompasses 657 million clinical notes and 1.3 billion lab tests. The company’s platform analyzes diverse healthcare data, including unstructured clinical notes, reports (echo, radiology, pathology), and structured data like lab tests, appointments, medications, diagnoses, and patient records. Additionally, nference can integrate specialized data modalities such as ECGs, genomic panels, and medical images (CT, MRI, PET).

Mayo Clinic and nference forge long-term strategic partnership

Mayo Clinic and nference have entered into a 12-year strategic relationship that involves Mayo Clinic Platform’s de-identified electronic health data and nference’s AI-driven nSights platform. The prestigious hospital wasn’t just a recent partner; they were an early investor in nference, involved in both the company’s Series B and C funding rounds, investing $20 million each at both stages.

Venky Soundararajan, Ph.D.

Venky Soundararajan, Ph.D.

“Mayo Clinic was involved from the start,” said Venky Soundararajan, Ph.D., chief scientific officer at nference. The prestigious hospital made a strategic investment when Dr. Gianrico Farrugia, the CEO of Mayo Clinic, announced the platform initiative in 2019.

“It’s pretty unusual for an academic medical center to make such a large investment,” Soundararajan said. One point of differentiation was the ability to go “above and beyond” HIPAA requirements

“We were competing with some of the tech giants exploring neural networks for various applications, including making physicians’ notes, which are unstructured text documents, fully HIPAA compliant and even exceeding HIPAA standards.”

Sniffing out COVID-19 with AI

The pandemic further highlighted the value of nference’s technology. Working with Mayo Clinic, nference’s AI models identified loss of smell (anosmia) and altered taste (dysgeusia) as early COVID-19 symptoms, even before positive PCR tests. This finding was significant enough to prompt the CDC to update their guidelines. “We also published two back-to-back papers in eLife, the peer-reviewed journal, demonstrating a potential causal link between early loss of smell or altered smell and COVID infection,” Soundararajan said.

Speaking of COVID, the pandemic underscored the need for rapid insights and collaboration between research and clinical care. “The walls between clinical research and clinical care had to come down and be replaced by bridges” to enable rapid discovery and knowledge sharing, emphasized Soundararajan. “There was a lot that we didn’t know about this pathogen that we had to learn about very, very quickly.”

Bridging the gap between data and discovery

This experience highlights the vast potential of not only unlocking previously inaccessible healthcare data with AI, but about contextualizing and uniting multiple data types to derive actionable information from multimodal medical data.


Maulik Nanavaty

Nference’s work in digital pathology through its Pramana initiative demonstrates a commitment to enhance research capabilities and inform therapeutic strategies. Anumana exemplifies nference’s targeted innovation in the diagnostic space. By focusing on extracting deeper insights from ECGs, a diagnostic tool that has been used for over a century, Anumana aims to unlock previously inaccessible information. As Anumana CEO Maulik Nanavaty explained, “We now have an approval for left ventricular ejection fraction signal; we have three of the breakthrough designations.” Current 510(k) clearances span everything from pulmonary hypertension to cardiac amyloidosis. Anumana has pharma collaborations with Novartis, J&J and Pfizer. (Anumana chief business officer Dave McMullin was featured in the inaugural episode of AI Meets Life Sci.)

Building a rich, multimodal data foundation

The key to unlocking the full potential of AI in healthcare lies in continuing to build out a deep, multimodal data layer and making that secure knowledge available for partners to extract valuable insights, Soundararajan said. While some companies may have more data on specific modalities, such as genomics or clinical codes, nference distinguishes itself by providing the most complete snapshot of patients possible while fully protecting their privacy.

As Soundararajan noted, “We don’t buy or sell data; we provide access for partners to deploy AI algorithms.” This approach enables partners to leverage diverse data types, from ECGs to lab tests. “You can imagine what the next two or three years of patients’ lives will be because you have an archetype for these patients. You can ask what interventions they should consider whether GLP-1s or conventional medicines, and so on and so forth,” Soundararajan said. “Lots of exciting opportunities in deidentified multimodal data for AI enablement that the parent company nference will continue to evolve.”

Beyond its work with Mayo Clinic and other academic medical centers, nference collaborates with biopharma companies that tap the company’s proprietary EMR-derived data, software products, and services to address challenges across the drug lifecycle, ranging from target discovery and clinical trial design to real-world outcomes research.