John Snow Labs Closes 2024 with Record Revenue, Customer Base Growth, and Open-Source Adoption, Thanks to State-of-the-art Healthcare-Specific Large Language Models

Rapid Innovation and Deep Industry Partnerships Contributed to 325% YoY Growth from Cloud Marketplaces, 30 New Research Collaborations, and 130M Spark NLP Downloads

John Snow Labs, the healthcare AI company, today announced record adoption of its open-source and healthcare-specific large language models (LLM) and natural language processing (NLP) solutions in 2024. This is the result of a concentrated effort to deeply integrate its technology across a range of cloud and data platforms, making it easier for customers to adopt and leverage its technology in a private, safe, and scalable way.

• Some of this year’s major achievements include: Making John Snow Labs’ AI models available in AWS , Azure , Databricks , Snowflake , and the Oracle Cloud Marketplaces The company reports a 325% year-over-year growth in revenue from these marketplaces.
• John Snow Labs has significantly expanded its work with academic medical centers and universities, jointly submitting grant applications with 30 new collaborators this year. The team was .
• John Snow Labs healthcare-specific LLMs were chosen as the only industry-specific LLMs available at launch on Amazon Bedrock Marketplace.

Customer Success

Against the hype cycle around Generative AI and “rush-to-pilot” across industries, John Snow Labs has focused on helping its customers get production-grade LLMs into  use, so  a real return on investment is achieved. This requires tackling challenges of privacy, security, integration, and scalability early on – crucial in a high-compliance industry like healthcare. Public case studies of successful initiatives published this year include:

• Roche Applies Healthcare-Specific LLMs to Build Oncology Patient Timelines and Recommend Clinical Guidelines
• WVU Medicine, a Health System with 25 hospitals, Maximizes Patient Care through AI-Enhanced HCC Code Discovery
• Functional Mind Leverages Medical Generative AI & Agents to Help Clinicians Provide Evidence-Based Care in Functional and Integrative Medicine

Peer-Reviewed, State-of-the-art Accuracy

John Snow Labs is relentlessly focused on providing the healthcare and life science ecosystem with the most accurate and capable AI models ever developed. Because state-of-the-art continues evolving at a breakneck pace, the team trains and tunes dozens of models each week, benchmarking and reproducing every new research innovation and novel technique. This involves daily collaboration between teams of data scientists and medical doctors to develop unique, high-quality datasets and benchmarks that correctly apply current medical knowledge. Outcomes from this effort include peer-reviewed papers establishing new state-of-the accuracy across the most common clinical language understanding tasks.

A sample of the company’s state-of-the-art excellence includes

• John Snow Labs’ “small” healthcare-specific LLM preferred 45%-92% more often on the dimensions of factuality, clinical relevance, and conciseness in a blind evaluation by practicing medical doctors versus GPT-4.
• John Snow Labs’ models making significantly fewer errors than the comparable AWS (50%), Azure (475%), and GCP (575%) services, respectively, while also outperforming ChatGPT by 33% on automated de-identification of a large real-world clinical text dataset.

On the task of question-answering about patient histories (Medical Text-to-SQL), John Snow Labs’ models outperformed GPT-4 by 15 points in logical form accuracy and by a wider margin open-source model based on Llama and Defog

Open-Source Contributions and Adoption

John Snow Labs continues to contribute to the open-source community with two teams focused on scalable language understanding (Spark NLP) and responsible AI (LangTest). Momentum in 2024 includes:

• Spark NLP adoption growing to 130 million total downloads-50 million more than the previous year.
• The curated Models Hub crossed 100,000 models, of which 63% are now LLMs. In contrast to community hubs like Hugging Face, this is an enterprise-focused hub to which only select models get accepted to after passing security, accuracy, and scalability tests.

Spark NLP now provides private batch inference for LLMs on commodity hardware, offering a more economical and faster way to process large amounts of LLM tasks.

“There are two things we promise our customers and the open-source ecosystem right now, said David Talby, CTO, John Snow Labs “The first is that we will keep you at the state-of-the-art, making sure access to the best healthcare AI ever created is always available. The second is that we’ll help you build solutions that are safe, effective, and production-grade – the ones you’ll benefit from after the AI bubbles burst.”