Posted on 01 November, 2023
In the ever-changing world of AI, few sectors hold as much promise as healthcare. Boston, in its pursuit of leveraging AI's transformative potential for healthcare solutions, embarked on an innovative journey. This voyage was made possible through a groundbreaking partnership with DKube, a One Convergence company, whose expertise reshaped AI development for Boston Medical.
Crafting AI algorithms: Two paths
The development of AI algorithms often presents a pivotal choice: one can opt for open-source tools, which offer flexibility, or embrace an MLOps framework for a structured, controlled approach. In healthcare, the importance of structure cannot be overstated due to stringent medical standards. Boston Medical, the healthcare arm of Boston Limited, recognised this necessity and set its sights on the structured path.
Boston Medical meets DKube
The transformational decision led Boston Medical to cross paths with DKube during the Supercomputing 21 (SC21) event in St. Louis, Missouri. After thorough presentations, live demonstrations and in-depth evaluation, Boston Medical made the strategic choice to transition their medical algorithm development to the DKube platform.
Reaping the rewards with DKube
Fast forward to 2023, and Boston Medical's decision to embrace DKube's expertise delivered transformative outcomes, enhancing crucial aspects of AI development:
- Data management simplified: DKube introduced a portal-like experience, streamlining data organisation, simplifying version control and maintaining a comprehensive inventory of data used for specific training.
- Streamlined environmental version control: Upgrading software with open-source tools had been a complex manual task. DKube introduced distinct software repositories, auditing capabilities and systematic metric capture for streamlined version control.
- Automated model version control: DKube automated model versioning, ensuring consistent metric production for every run and accommodating custom metrics, simplifying change tracking.
- Efficient parameter and artifact management: With DKube, data remained orderly and accessible through an intuitive management interface, streamlining data retrieval, expediting insights from past experiments and facilitating swift model updates.
- Enhanced data drift monitoring: DKube introduced configurable Alert Settings for data drift, threshold limits for monitoring and the ability to assess model latency and accuracy. This ensured performance issues were promptly identified and addressed.
- Scalability realised: DKube brought relief to the management of Boston Medical's extensive portfolio of models.
Efficiency and support
Beyond the transformative benefits, Boston Medical highlighted DKube's efficiency and support as vital aspects of their partnership. In the healthcare sector, where security and compliance are paramount, DKube played a crucial role in addressing these critical concerns.
Boston Medical's journey with DKube, from initial engagement to their collaborative efforts, now empowers them to execute end-to-end processes efficiently, from running inferences to testing models with new data. With a focus on visualising results and collaboration through TensorBoard, Boston Medical shares outcomes seamlessly with key stakeholders, enhancing healthcare outcomes and solidifying their partnership with DKube as a critical step in their AI journey.
This collaboration redefines the landscape of healthcare AI, making it structured, efficient and accessible for customers, embodying the commitment of Boston and DKube to advancing AI for the betterment of healthcare and beyond.