How AI experts and scientists are collaborating to drive positive health outcomes
Artificial intelligence is accelerating research and clinical practice, but the biggest gains happen when scientists, engineers, and clinicians solve problems together. At UT San Antonio's MATRIX AI Consortium for Human Well-being, that collaboration is the point. The team is pairing domain expertise with practical machine learning to improve care and educate the next wave of AI practitioners.
In a recent Launchpad podcast episode, UT San Antonio's John Elizondo hosts a conversation with three MATRIX leaders: Amina Qutub, PhD, and Mark Goldberg, MD, who co-lead the thrust on augmenting human capabilities, and Amanda Fernandez, PhD, who leads machine learning and deployment. Their focus: building AI that is secure, trustworthy, and useful at the bedside and the bench.
Inside the conversation
The discussion moves from student career advice to concrete use cases in trauma care and stroke management. The through line is simple: rigorous science, responsible AI, and solutions that fit into real clinical workflows.
What matters to researchers
- Trust and safety first: Security, privacy, and reliability are non-negotiable in clinical and research settings. Frameworks like the NIST AI Risk Management Framework help teams set standards from day one.
- Data that stands up to scrutiny: Clear provenance, representative cohorts, and unbiased labeling are essential. Reproducibility and external validation should be routine, not an afterthought.
- Human-in-the-loop: AI should augment clinical judgment, not replace it. Interfaces must fit the way clinicians think and work under pressure.
- Clear translation pathways: From hypothesis to deployment, teams need defined steps for testing, regulatory review, monitoring, and post-deployment auditing.
Practical applications discussed
- Trauma care: Early-warning tools that analyze vitals and EHR streams to flag deterioration sooner. Decision support that prioritizes imaging or specialist consults when seconds matter.
- Stroke management: Models that help identify large-vessel occlusions on imaging, cut door-to-needle time, and predict risks like hemorrhagic conversion. Rehabilitation planning that adapts intensity based on recovery signals.
- Operational gains: Triage support, bed and resource allocation, and follow-up planning that reduces avoidable delays and improves continuity of care.
Advice for students and early-career scientists
- Build end to end: Take a small clinical question from problem framing to data prep, modeling, validation, and a simple interface. You'll learn fast where the real friction is.
- Know the science and the context: Stats, signal processing, and causal inference matter. So do clinical guidelines, workflow constraints, and equity concerns.
- Document everything: Data sources, consent, labeling instructions, and known biases. Good documentation beats a marginal lift in model accuracy.
- Prioritize interpretability: Use methods that reveal why a prediction was made. Clinicians need clarity, not mystery.
- Collaborate early: Pair with a clinician, data steward, and security lead from the start. You'll avoid rework and design for real use.
- Ship responsibly: Build privacy, access controls, and monitoring into the first prototype. Treat model updates like any high-stakes software deployment.
Why this approach works
Mixed teams catch blind spots. Clinical leaders surface edge cases; scientists keep methods sound; engineers make systems usable and maintainable. That blend turns promising models into tools that improve outcomes and hold up under peer review and real-world constraints.
Listen and follow
Launchpad, produced by the Office of University Strategic Communications at UT San Antonio, features experts applying AI to meaningful problems. You can find episodes on Apple, Spotify, SoundCloud, and YouTube. Future conversations with UT San Antonio researchers are on the way-subscribe so you don't miss practical insights you can apply in lab, clinic, or translational projects.
Explore further
- For structured learning paths aligned to specific roles, explore AI course paths by job.
- For guidance on trustworthy AI, review the NIST AI Risk Management Framework.
Your membership also unlocks:
