AI Experimental Systems Research Scientist (Causal Learning & Adaptive Experimentation)

Collaborate with Innovative 3Mers Around the World

Choosing where to start and grow your career has a major impact on your professional and personal life, so it's equally important you know that the company that you choose to work at, and its leaders, will support and guide you. With a wide variety of people, global locations, technologies and products, 3M is a place where you can collaborate with other curious, creative 3Mers.

This position provides an opportunity to transition from other private, public, government or military experience to a 3M career.

The Impact You'll Make in this Role

As an AI Experimental Systems Research Scientist in 3M's Corporate R&D organization, you will work on a small, deeply technical team developing foundational methods for always-on learning systems that reason, experiment, and adapt in complex, non-stationary environments. This role exists because learning systems that do not actively preserve identifiability, causal validity, and epistemic calibration fail in principle—not merely in performance.

You will collaborate closely with researchers across statistics, cognitive science, and machine learning to design systems in which experimentation, inference, and uncertainty are first-class components of the learning process itself. This is not a conventional data science or applied machine learning role. The work focuses on how learning systems must structure experiments, manage interference and delayed effects, govern their own representations, and remain epistemically correct over time.

This role is well suited for someone who enjoys working from first principles, designing rigorous experimental machinery, and translating statistical theory into systems that operate continuously in the real world.

Here, you will make an impact by:

• Designing and implementing adaptive experimental systems that operate continuously under nonstationarity, interference, and delayed or indirect outcomes.
• Developing causal estimands, randomization schemes, and inference procedures whose primary goal is identifiability and validity, not just reward optimization.
• Embedding rigorous experimental control directly into learning systems, including experimentation on the system's own learning mechanisms, parameters, and representational choices.
• Translating principles from experimental design, causal inference, and sequential decision-making into robust, always-on system behavior.
• Working from whiteboards, research discussions, and evolving specifications—not fixed product requirements or static datasets.
• Implementing and maintaining research code that supports hierarchical experimentation, baseline control streams, and statistically valid online inference.
• Creating diagnostics, monitoring tools, and guardrails to ensure learning systems remain calibrated and do not stabilize spurious structure over time.
• Collaborating with interdisciplinary researchers to stress-test experimental learning mechanisms under realistic, adversarial conditions.

Your Skills and Expertise

To set you up for success in this role from day one, 3M requires (at a minimum) the following qualifications:

• Ph.D. in Statistics, Biostatistics, Economics, Computer Science, Data Science, Operations Research, or a closely related field (completed and verified prior to start).
• Deep grounding in experimental design and statistical inference, including randomized experiments and causal estimands.
• Demonstrated ability to implement research-grade statistical or experimental methods in a general-purpose programming language (e.g., Python).
• Experience working in research settings where the problem definition evolves and correctness takes precedence over convenience.

Additional qualifications that could help you succeed even further in this role include:

• Experience with adaptive or sequential experimentation (e.g., response-adaptive trials, causal bandits, best-arm identification).
• Familiarity with causal inference frameworks spanning both design-based and model-based approaches.
• Strong intuition for identifiability, bias–variance tradeoffs, and statistical validity in complex, real-world settings.
• Experience working with nonstationary systems, concept drift, or delayed feedback loops.
• Experience reasoning about interference, carryover effects, time-varying treatments, or non-independent experimental units.
• Comfort designing experiments where the learning process itself is the object under experimental control.
• Familiarity with hierarchical or clustered experimental designs and multi-level inference.
• Interest in foundational questions about how autonomous systems should reason, experiment, and adapt in the world.
• Ability to communicate complex statistical ideas clearly to interdisciplinary collaborators.
• Curiosity, intellectual humility, and a strong preference for epistemic correctness over s