Apply

Fellows accepted into the program will gain insight into research in both academia and industry. Our prestigious fellowships are fully funded for a three-year period and provide the fellows’ salaries, fringe benefits, and a generous travel allowance. Our fellowships are open to outstanding candidates seeking career advancement in biomedical data sciences.

The Yale-BI Joint Steering Committee (JSC) has established the following data-driven research theme for the 2026 program. The steering committee will consider how well proposed research projects align with the prioritized theme when judging submissions and postdoctoral applications. Selected applications will be further reviewed for approval.

Proposal of Research Themes for the Yale-Boehringer Biomedical Data Science Fellowship 2026

The Yale-BI Biomedical Data Science Fellowship Program for 2026 invites innovative research proposals that harness cutting-edge AI to tackle critical challenges in drug development. This initiative combines Yale’s leadership in AI and Foundation Models (FM) with Boehringer Ingelheim’s expertise in translational science to accelerate discovery and patient impact.

The 2026 program focuses on three transformative areas designed to drive innovation at the intersection of computational biology, AI, and drug discovery:

1. Foundation Models for Patient Stratification: Enabling precision medicine by integrating multi-omics, Real-World Data (RWD), and clinical data.
2. Foundation Models for Safety Prediction: Addressing drug development bottlenecks by improving early toxicity assessment.
3. Agentic AI: Leveraging autonomous agents to generate scientific hypotheses and recommend subsequent experiments.

These themes aim to transform treatment strategies for chronic liver and kidney diseases, inflammatory diseases, and cancer.

The Challenge: A major hurdle in drug development is the lack of comprehensive, data-driven insights to guide patient selection across the translational pipeline. Traditional approaches often rely on operational criteria which, while necessary for logistics, fail to fully capture the complexity of diverse patient populations. While current methods have seen success in cancer and genetic mutation-based rare diseases, they have had limited success in other therapeutic areas.

The Goal: This project seeks to develop foundation model-based methodologies that synthesize multi-omics, real-world evidence, and clinical data to generate actionable insights for intelligent patient segmentation and risk stratification.

Key Focus: Proposals should focus on developing novel methods that move beyond diseases driven by single-gene mutations. The objective is to enable broadly applicable strategies that segment patient populations and disease progression to inform precision therapeutics.

The Challenge: Drug development faces high attrition rates due to safety failures in early clinical trials, largely driven by the poor translatability of preclinical toxicity findings to humans.

The Goal: This project aims to develop foundation-model-based approaches to quantify the likelihood of unmanageable human toxicity early in the preclinical pipeline—potentially before the synthesis of a molecule—to reduce the risk of costly failures later in the clinic.

Key Focus: Research should leverage recent technological advances in AI foundation models and utilize a wide range of data sources and modalities to learn complex biological relationships for toxicity assessment.

The Challenge: Standard research methodologies often face prolonged timelines in hypothesis testing and experimental planning.

The Goal: Agentic AI systems, powered by Large Language Models (LLMs) and multi-agent architectures, have the potential to redefine scientific research by automating hypothesis generation, experimental design, and execution. These systems can close the loop of the scientific method, significantly reducing preclinical timelines and enabling novel approaches to trial optimization.

Key Focus: This project will develop multi-agent AI frameworks to autonomously generate scientific hypotheses that enhance disease understanding and identify novel therapeutic targets by demonstrating a strong Target to Disease Link (T2DL). It will leverage advanced agentic architectures, such as "AI co-scientist" prototypes, to integrate multi-omics, in vitro, in vivo, and clinical data for experimental automation.

Potential Applications Include:

• Autonomous Genetic Perturbation Design: Accelerating disease mechanism studies and therapeutic target identification.
• Novel Target Identification and Prioritization: Synthesizing heterogeneous data to speed up upstream drug discovery.