League launches League Labs to advance healthcare AI

Today, we announced the formation of League Labs^TM, an applied research group advancing the state of healthcare AI. League Labs is dedicated to publishing research, benchmarks, and open-source datasets to drive innovation in healthcare, with its first release available this month.

League has spent over a decade building AI infrastructure for healthcare, accumulating billions of interactions with consumers and patients across every demographic, through payers and providers across the US, Canada, and global markets. That depth, spanning care navigation, clinical signal processing, and personalized health engagement at scale, is the foundation League Labs^TM is built on. The group’s mission: to build League’s intelligence layer, powering agentic applications in healthcare, and a foundation the entire healthcare AI community can build on.

League Labs^TM is pursuing three initiatives:

• League SYNTH, a platform that generates Privacy Preserving Personas, richly detailed synthetic patients that mirror the diversity and clinical complexity of real populations, modelled from population health statistics, without exposing a single medical record. SYNTH enables developers and researchers to align their work to how people actually interact with the healthcare system, ensuring every population segment is represented during model training. 

• A family of small language models (SLMs) purpose built and tuned for agentic healthcare applications, and designed to outperform general purpose models on cost, latency, and clinical accuracy in key patient and member facing areas. An early example is League’s Guardrails model that keeps member-facing conversations safe, on-policy, and clinically appropriate in real time.

A set of patient-centered benchmarks that evaluate AI performance across the full health journey: from understanding options and navigating care to identifying next best actions and ensuring support that is safe, timely, and clinically appropriate. This includes League SYNAPSE, a benchmark that measures an agent’s ability to recognize health status indicators in natural human-agent conversations, determine which clinical workflows or health data representations should be updated, and consolidate those signals into agentic memory systems that improve downstream care support.

League Labs^TM operates with access to NVIDIA H200 GPUs and is working with leading academic AI institutes, including the Vector Institute, a leading research organization that transforms cutting edge artificial intelligence research into practical solutions.

“General healthcare intelligence (GHI) is the next frontier of AI, and the talent and clinical depth to lead it exists in the global research community,” said Michael Serbinis, League CEO and co-founder. “League Labs^TM is our commitment to building that frontier together: leveraging over a decade of healthcare intelligence to create an AI foundation the entire industry can build on, openly, responsibly, and with the rigor that healthcare demands.” 

Today, League Labs^TM is announcing that we will publish our first SYNTH-generated dataset on Hugging Face this month. The v1 release targets approximately 1,000 synthetic patient samples across multiple health domains, drawn from general population health statistics, sized to match established benchmarks like OpenAI’s HealthBench Hard while keeping evaluation accessible to independent and academic researchers.  The SYNTH-generated dataset targets critical areas of agentic care that currently do not have a standard benchmark and reflects the real-world distribution of health conditions. The data is generated using leading open foundation models including NVIDIA NeMo, Google Gemma 4 and Cohere Command A+, on training infrastructure through the TELUS Sovereign AI Factory. 

“We believe the future of healthcare AI will be built on open architectures and hybrid systems that combine the broad capabilities of frontier models with smaller, purpose built models tuned for healthcare.” said Aiden Arnold, VP of AI Models at League and head of League Labs^TM. “Our focus is the intelligence layer that makes AI clinically relevant, operationally efficient, and safe enough for real healthcare workflows. General healthcare intelligence is one of the most important opportunities AI has in front of it. We have the chance to deliver better outcomes, extend healthspans, and unlock the pathways that get people the right care at the moment they need it. League Labs^TM is how we make sure that future actually gets built.”

About League

Founded in 2014, League is the leading healthcare consumer experience (CX) platform, powered by artificial intelligence (AI), reaching more than 70 million people around the world and delivering the highest level of personalization in the industry. Payers, providers, and healthcare innovators build on League’s platform to deliver high-engagement healthcare solutions proven to improve health outcomes. League has raised over $285 million in funding to date, powering the digital experiences for some of healthcare’s most trusted brands, including HCSC, Santa Clara Family Health Plan, SCAN and Manulife. For more information, visit league.com.