Product descriptions:
Children’s National Hospital, based in Washington, D.C., was established in 1870 to help every child grow up stronger. Today, it is one of the top 10 children’s hospitals in the nation and ranked in all specialties evaluated by U.S. News & World Report. Children’s National is transforming care on Databricks. By utilizing the Databricks Data Intelligence Platform, Children’s National reduced model training time from months to minutes, deployed GenAI agentic tools like Spotlight AI to streamline clinical workflows and hospital accreditation, sped up nurse onboarding and scaled predictive analytics for critical care. As a leader in research and innovation, Children’s National is building domain-specific GenAI applications on Databricks — securely, cost-effectively and at scale — to deliver faster insights and better outcomes for children with complex healthcare needs.
Transforming pediatric care with data intelligence
For the fifteenth straight year, Children’s National ranked in 10 specialty services and is the highest U.S. News ranked children’s hospital in Washington, D.C., Maryland and Virginia.
Yet, Children’s National struggled with a fractured data architecture that created barriers to basic reporting, timely insights and advanced research. “Historically, Children’s National had fractured data architectures and challenges with getting basic insights and analytics promptly,” says Mike McLaughlin, Head of AI/ML and Advanced Analytics at Children’s National.
Supporting clinical workflows was challenging, particularly when onboarding new staff who required quick access to policies and procedures. Siloed information slowed their integration and affected the delivery of timely care. “We hire many nurses, and each organization has unique policies and practices,” Mike explains. Even experienced employees encountered difficulty finding essential information, which had implications during accreditation surveys.
At the same time, Children’s National researchers were constrained to training models on individual laptops. Resource limits forced them to work with small samples, making it impossible to scale their efforts. “Teams were stuck running models on laptops, pressing run and waiting days, only to find out the results weren’t usable. And because everyone was working on their own machines, the process wasn’t reproducible and collaboration was nearly impossible,” says McLaughlin. The hospital required a fully modernized analytics and AI platform that could unify data and enable teams to deliver faster, more effective outcomes.
Building an AI-powered foundation on Databricks
Children’s National used the Databricks platform as part of its efforts to unify its data, accelerate research, and build GenAI solutions into clinical and operational workflows. One notable innovation is Spotlight AI, a retrieval augmented generation (RAG) agent that centralizes organizational policy materials and critical nursing policies.
“We connected this agent to these core policy and procedure knowledge bases to help with onboarding and accreditation activities. Staff would report that they don’t know, or understand, our policies and how we do things at Children’s. Now, they can just ask Spotlight AI the exact question and the chatbot shares the answer with a link to the source document,” says McLaughlin. Spotlight AI not only saves time and improves onboarding experience, but also helps ensure high degrees of compliance with strict regulatory standards
In addition to knowledge management initiatives, the hospital is implementing predictive analytics for staffing forecasts, workforce optimization, and patient throughput. Artificial intelligence models support leadership in predicting patient volumes and determining nursing requirements up to six months in advance. Another system undergoing validation is the Criticality Index, a machine learning-based tool designed to proactively identify pediatric patients at risk of clinical deterioration during their hospital stay. Developed by Children’s National researchers, this system seeks to enhance quality of care by enabling clinicians to detect early warning signs of pediatric decline, thereby facilitating timely intervention and appropriate escalation to critical care units when necessary.
Children's National's AI/ML team leverages Unity Catalog for data and model governance, ensuring consistent and secure collaboration across teams, so they can scale AI initiatives with confidence and directly improve patient care outcomes. MLflow gives Children’s National a single system to track, evaluate and manage both predictive models and GenAI apps. The team utilizes it to version datasets and models for projects such as the Criticality Index, ensuring reproducibility and compliance. For GenAI tools such as Spotlight AI, MLflow supports prompt and model evaluation, allowing outputs to be compared and safely integrated into production.
Mosaic AI Gateway gives Children's National a secure and scalable way to deploy and manage AI models. Through a central interface for proprietary, open source and custom models, it makes integration easier and speeds up innovation while maintaining strong governance and security. With Mosaic Vector Search, they can connect LLMs to their internal knowledge bases. This ensures staff receive precise answers grounded in authoritative sources through Spotlight AI. “The fact that it’s a unified experience doesn’t matter if I’m managing an AI agent or a traditional model. This has been great for the developer experience,” says McLaughlin. Partnerships have also been crucial. Slalom Consulting supported the build on Databricks and the development of GenAI use cases. This rapid scale-up reflects how Databricks has grown in tandem with the hospital’s needs.
Measurable impact and a culture of data-driven innovation
While some initiatives are still in validation stages, Databricks’ impact has already been significant across Children’s National. Staff report faster time to find information with real-time responses powered by Spotlight AI. “It’s hard to quantify because we had no baseline, but there’s a palpable excitement around data that just didn’t exist before,” says McLaughlin.
The data intelligence platform has also energized the hospital’s research community, enabling them to utilize real-world data and push the boundaries of medical knowledge, ultimately changing how the hospital delivers care holistically. Residents and scientists are writing new grant proposals, cohort identification for clinical research is accelerating, and ideas once considered impossible, such as patient-friendly rounding notes transformed by GenAI, are coming to life. “’My goodness, there’s hope for how we can do data and AI work at this organization now,’” says McLaughlin. “That statement has come out of stakeholders again and again.”
Operationally, model training timelines have been reduced from months to minutes. Faster model training means clinicians get actionable insights sooner, improving patient care. For example, tasks that previously took 30–45 minutes per day to review documents are now instantaneous with the Spotlight AI chatbot, helping staff access the answers they need immediately-reducing the risk of errors and ensuring high accreditation standards.
“Every time we showcase these capabilities, folks are blown away by what’s possible,” says McLaughlin. “For once, we can say yes to requests that used to be impossible. And that’s changing the way our entire organization thinks about data.”