Drew Prinster

Drew Prinster

Hi, I’m Drew. I’m a PhD student in computer science at Johns Hopkins University. I aim to advance our understanding and the realization of regulatable AI systems, especially in healthcare and biomedical research. I approach this goal with the following agenda:

Diagram of my research agenda for regulatable AI: statistical uncertainty quantification under distribution shifts as the foundation, supporting pillars of controllable AI, monitorable AI, and human-AI teaming / AI for health.
  1. Statistical Uncertainty Quantification: I contribute to a foundation for AI uncertainty quantificiation, for example by extending conformal prediction to arbitrary data shifts and those induced by AI agents.
  2. AI Control: Building on this bedrock, I work on methods to provably enforce stakeholders’ safety constraints on AI agent behaviors, even as they advance in capabilities.
  3. AI Monitoring: To audit conditions required for safety assurances, I develop sequential hypothesis testing tools for system monitoring.
  4. Human-AI Teaming: To better understand the constraints to regulate for in the first place, I collaborate with stakeholders in healthcare and biomedical research, and I study how to design systems for effective human-AI teams.

I am advised by Professor Suchi Saria and Professor Anqi Liu. In 2025, I interned with Prescient Design at Genentech, where I began developing conformal policy control. I have also worked closely with Professor Chien-Ming Huang.

Previously at Yale University, I completed my B.S. (with distinction) in computer science and mathematics and was a Yale Global Health Scholar. At Yale I contributed to global health volunteering and research with the nonprofit Unite For Sight; to microbiome, mental health, and social-network research with Dr. Nicholas Christakis at the Yale Human Nature Lab; and to statistical testing of metagenome alignments with Dr. Mark Gerstein in the Gerstein Lab.

Contact: drew [at] cs [dot] jhu [dot] edu

Selected Publications

AI Control: Safety Constraints for AI Agents

Conformal Policy Control
  • AI Agents
  • Risk Control
  • Conformal Prediction
  • Distribution Shift
  • Healthcare Application
  • Bioscience Application

AI Monitoring: Post-Deployment Safety Monitoring

WATCH: Adaptive Monitoring for AI Deployments via Weighted-Conformal Martingales
  • Monitoring & Sequential Testing
  • Distribution Shift
  • Conformal Prediction
  • Healthcare Application
E-valuator: Reliable Agent Verifiers with Sequential Hypothesis Testing
  • AI Agents
  • Monitoring & Sequential Testing
  • Healthcare Application

Uncertainty Quantification under Distribution Shift

Conformal Validity Guarantees Exist for Any Data Distribution (and How to Find Them)
  • Conformal Prediction
  • Distribution Shift
  • AI Agents
  • Bioscience Application
JAWS-X: Addressing Efficiency Bottlenecks of Conformal Prediction Under Standard and Feedback Covariate Shift
  • Conformal Prediction
  • Distribution Shift
  • Efficient Uncertainty Quantification
  • Bioscience Application
JAWS: Auditing Predictive Uncertainty Under Covariate Shift
  • Conformal Prediction
  • Distribution Shift
  • Efficient Uncertainty Quantification

Human-AI Teaming: Designing AI for Effective Collaboration

Care to Explain? AI Explanation Types Differentially Impact Chest Radiograph Diagnostic Performance and Physician Trust in AI
  • Human-AI Interaction
  • Explainability/Interpretability
  • Healthcare Application