Jill Goldsberry, Go-To-Market AI Senior Solutions Consultant at Centaur Labs, is in conversation with Host Simon Hodgkins on this episode of VistaTalks. The conversation explored how AI is transforming healthcare, why high-quality labeled data is the foundation of reliable models, and how Centaur Labs is helping organizations overcome some of the most pressing challenges in this space.
Connecting Healthcare and AI
Jill’s role at Centaur Labs centers on bridging healthcare and life sciences organizations with the expert data resources they need to build reliable AI systems. From annotating medical images for diagnostic models to labeling biosignals, clinical text, or surgical videos, her team ensures datasets are expertly crafted, regulatory-ready, and delivered at scale.
As Jill explained, healthcare AI development is not just about building models. It requires data that is both accurate and carefully vetted. “Annotation is often treated as a simple mechanical task, but in reality, especially in healthcare, it requires domain expertise and rigorous quality controls,” she said.
Integration Without Disruption
Healthcare workflows can be complex and highly specialized; yet, Centaur Labs has designed its platform to integrate seamlessly into these systems. Instead of forcing teams to adopt new processes, Centaur adapts to existing workflows, whether that involves model training, validation, or post-market monitoring. This flexibility enables healthcare organizations to enhance their existing operations without unnecessary disruption.
The Power of Expertise at Scale
One of the most fascinating aspects of the conversation was Jill’s explanation of how Centaur Labs leverages a global community of over 40,000 contributors, many of whom are clinicians, researchers, or healthcare professionals in training. Through a mobile platform with gamification elements, contributors complete labeling tasks in an engaging way, wherever they are.
This approach combines scale with expertise. Consensus algorithms aggregate multiple expert judgments into high-confidence labels, ensuring quality remains high even at speed. The result is a dynamic, flexible system that delivers both accuracy and efficiency.
Real-World Use Cases
Jill highlighted several compelling projects, including work with Eight Sleep on snore detection models that enhance sleep tracking and may aid in identifying sleep-related breathing disorders. Another example was Centaur’s collaboration with Microsoft Research to create bilingual radiology datasets. These datasets are crucial for training diagnostic models that operate across languages, thereby expanding access to reliable AI-driven tools for a broader global audience.
Tackling the Toughest Data Challenges
Some of the most challenging domains for annotation include surgical video, ECGs, and lung sounds, which require both technical interpretation and clinical context. To address this, Centaur relies on highly trained annotators, rigorous training, and multiple reviewers per data point, alongside automated checks and statistical controls. This ensures consistency and accuracy, even when working with the most complex data types.
Building Trust Through Compliance
Healthcare AI must meet the highest standards of trust and compliance. Jill emphasized that Centaur treats regulatory readiness as a requirement from the start, not an afterthought. From secure data handling to audit trails and contributor vetting, every dataset is designed to meet the evidentiary standards required for FDA submissions or CE marking. This approach ensures that data is ready for both pre-market and post-market use cases.
“Everything is moving faster,” Jill observed, “and we’re building infrastructure that supports these ongoing data pipelines so expert review can happen closer to the point of decision making.”
Emerging Trends
When asked about emerging trends, Jill noted a shift from static, one-time datasets toward dynamic, continuously updated ones that evolve alongside machine learning models. She also pointed to greater collaboration between human experts and AI, where annotators refine edge cases and improve model explainability.