A recent survey by CIO.com found that while 80 percent of companies are considering new AI projects and nearly 90 percent plan to invest in AI technology, only 17 percent of these companies have moved their projects into production. Translating science projects—usually in the form of notebooks or proof-of-concept—into revenue-generating, enterprise-grade products and services is still a high bar for most companies. There are numerous challenges across all areas of the AI development lifecycle. These range from data integration, quality, privacy, governance, and enrichment challenges, to model deployment hindrances that include support for model governance, versioned repositories, auto-scaling, monitoring, roll-out and roll-back, and audit. In addition, issues in model monitoring for concept drift, bias, and online accuracy as well as operational aspects like security controls, user and access management, scalability, monitoring, disaster recovery, and software upgrades further make it difficult for enterprises to get their AI to production quickly. In short, there is a whole new AI enterprise architecture to put together, which most companies do not realize or build over in their first few projects. Many teams instead choose to start without any plan and build their AI platform internally over time. This works when the first AI projects are small-scale and not business critical. However, it is not an option when the system must be production-grade or when working within a high compliance industry like healthcare or life science.
An alternative is proposed by the major public cloud providers who market ‘machine learning-as-a-service.’ Cloud services also sport strong support for scalability, elasticity, and centrally-managed access control. On the other hand, this approach requires complete lock-in to a specific cloud provider—with very limited customization or integration with existing software architecture. Such solutions have also been criticized as prohibitively expensive once the volume of either data, models, inference, or team size starts growing.
There is also a wide range of point solutions for specific parts of the AI platform architecture: products for model training, model deployment, model governance, and others. However, these tools actually make the AI platform architecture more complex—adding another piece that must be integrated in terms of both technology and development process. The fundamental goal of an AI platform is to provide that unified architecture and answer one question: How do I get real AI projects from concept to production as fast as possible?
A Proven End-To-End Solution
John Snow Labs has grown to be the most prominent pure-play AI company in the healthcare and life science space. Today, the company’s AI platform has been used by some of the world’s largest firms and has been publicly presented and vetted in leading technical AI and data conferences. For instance, Kaiser Permanente, one of the U.S. largest health plans, serving 12.3 million patients across 39 hospitals with over 220,000 employees, uses John Snow Lab’s AI platform to build its enterprise data science platform and apply it to optimize patient flow in hospitals.
Many companies claim to have an enterprise-grade, high-compliance, unified AI platform. Only John Snow Labs has the customer success and technical depth to prove it
That being said, John Snow Labs’ AI platform has also been vetted for delivering the security, privacy, and compliance requirements necessary to process protected health information (PHI). The controls it implements have been mapped to show coverage against multiple regulatory frameworks, including HIPAA, FISMA, and FERPA.
Cloud Benefits, Without the Lock-In
As the above component diagram shows, John Snow Labs’ AI platform provides an integrated toolset that covers all stages of the AI development lifecycle. All of the tools are deployed and operated together and are managed via a unified identity and single sign-on system, monitoring, and cluster management. As one would expect from a true platform, it not only provides a toolset but also embeds a set of best practices on how to use them. It does so by clearly defining the five roles that participate in the AI lifecycle: data engineers, data analysts, data scientists, application engineers, and DevOps engineers.
As a result of the unique set of capabilities provided by its AI platform, today Jon Snow Labs has attracted a combination of Fortune 1000 companies who use it as the basis for their own enterprise AI platform (by branding, extending, and integrating it into the rest of their software architecture) as well as small companies and start-ups who need to get to market fast with a proven, end-to-end solution. The AI platform can be deployed on a new cluster in as little as two hours, although John Snow Labs bundles it with a longer onboarding session that includes training, personalized design, and architecture sessions.
John Snow Labs also offers turnkey services that combine licensing its AI Platform with custom development of machine learning, deep learning, or natural language processing models that address the needs of specific customers. Such projects are usually done by a joint team that combines data scientists and engineers from John Snow Labs and the customer teams, so that in addition to getting to market quickly, the customer’s team gains hands-on expertise and on-the-job training in using the platform on a daily basis. This was the approach taken by Usermind, who built a new AI revenue stream from scratch in three months using our AI platform.
An Innovative Game Plan for the Future
Over the last four years, John Snow Labs has successfully assisted companies in realizing the potential of data science, proving its unique value proposition. Going forward, the company has an aggressive roadmap that will keep its customers at the forefront of the AI ecosystem: adding support for automated experiment tracking, centralized feedback collection, and model retrain pipelines; automated model development (Auto ML) and integration with additional best-of-breed tools are coming soon. The platform currently also includes the most recent versions of the award-winning Spark NLP and Spark NLP for Healthcare libraries as well as turnkey integration with reference datasets available through John Snow Labs’ frictionless data market. “Our customers consider us not just as a software provider but also as a managed service that continuously tracks the fast-evolving AI landscape and adds the best innovations to their toolbox. Every quarter, the definition of an AI platform keeps expanding and our product keeps up with it,” explains Ali Naqvi, AI Platform product manager at John Snow Labs.
John Snow Labs firmly believes that AI will have a tremendous impact on the healthcare and life science industries. With this strong conviction, the company is all poised to break new grounds and help organizations realize the potential of AI in the years to come.