Xu Lab · How we research, learn, and grow
Our Culture & Principles
We seek to create new capabilities for understanding and discovery, and to help people grow into researchers who can ask important questions, think independently, and build work that others can carry forward.
Read the full Lab Charter01 · Purpose
Creating New Capabilities for Understanding and Discovery
We investigate important scientific questions and explore more powerful ways of conducting research: how to represent problems, organize information, generate evidence, learn underlying patterns, and turn understanding into new designs and actions.
We are interested in questions that remain insufficiently understood yet may have lasting significance. We examine the questions themselves, including their assumptions, objective functions, search spaces, and evaluation criteria. We want discoveries, methods, and systems to support further work and help others make reliable judgments.
Read the corresponding charter section02 · Learning
Nurturing Curiosity, Learning to Think Independently, and Growing into Independent Research
Students may arrive with interests that are not yet clear and questions that are still taking shape. Through reading, discussion, and small exploratory studies, we help them discover what they want to understand and develop questions they want to keep pursuing.
We make thinking itself part of shared learning. Mentors and students explain how they chose an approach, what evidence supports a conclusion, and what would change their minds. Answers from papers, experts, the PI, peers, and AI all remain open to reasoned examination. Independence also includes recognizing limitations, seeking help, and collaborating well.
Students can focus on one or two of twelve research roles, from framing questions and designing learning objectives to building scientific AI systems and research infrastructure. The role map connects these capabilities to concrete career directions and examples of research contributions, with development goals agreed together.
Read the corresponding charter sectionStudent development
Who You Can Become
Through real research, we help students develop the ability to frame important questions, design learning and exploration methods, gather and judge evidence, understand scientific mechanisms, and build systems and tools for sustained discovery.
Each student can develop a combination of strengths that reflects their interests, background and career goals, and adjust it as they gain experience. Independent thinking, curiosity, research integrity, collaboration and continued learning are foundations for every role.
Research roles we emphasize
Scientific Problem Framer
Identify important unknowns and turn broad interests into worthwhile, testable research questions.
Example contribution: Scientific Problem Framer
A problem brief that sets out current knowledge, competing explanations, the key gap, feasible tests and conditions for revising the question.
Back to career directionsLearning Objective Designer
Design learning signals, optimization objectives and constraints; decide what a system should pursue.
Example contribution: Learning Objective Designer
A set of objectives and constraints, with the rationale, comparisons and situations in which the design may fail.
Back to career directionsSearch Space Designer
Define the variables, candidates and combinations to explore, including their feasible ranges and boundaries.
Example contribution: Search Space Designer
An executable space of molecules, materials, formulations or experimental conditions, explaining what is excluded and the consequences.
Back to career directionsScientific Representation Designer
Choose variables, coordinates, scales and representations that make key relationships easier to understand and learn.
Example contribution: Scientific Representation Designer
A representation compared across settings, showing which information it preserves, compresses or loses.
Back to career directionsModel Architecture Designer
Design how information interacts, propagates, aggregates and updates within a model.
Example contribution: Model Architecture Designer
A model prototype and architectural comparisons that explain why each component is needed and where the design fails.
Back to career directionsEvidence and Experiment Designer
Decide which observation, simulation or experiment would distinguish explanations and reduce a key uncertainty.
Example contribution: Evidence and Experiment Designer
An evidence-gathering plan that identifies possible results, alternative explanations and how each result would change the next step.
Back to career directionsScientific Evaluation Designer
Establish metrics, baselines, controls and tests to judge whether results represent real scientific progress.
Example contribution: Scientific Evaluation Designer
A reusable evaluation protocol that can detect superficial gains, shortcuts and conclusions that go beyond the evidence.
Back to career directionsScientific AI Systems Architect
Organize models, data, simulations, experiments and human judgment into a coordinated research process.
Example contribution: Scientific AI Systems Architect
A working discovery or design workflow that shows how each round updates understanding and subsequent actions.
Back to career directionsAI Research Infrastructure Designer
Build the data, computing, training, tracking and reproducibility environment that supports sustained research.
Example contribution: AI Research Infrastructure Designer
A reusable research platform with repeatable workflows, documentation and clear maintenance responsibilities.
Back to career directions
Further research strengths to develop
Mechanistic Researcher
Connect theory, computation and experiment to develop mechanistic explanations that can be tested.
Example contribution: Mechanistic Researcher
A mechanistic argument with a traceable chain of evidence, unresolved alternatives and a clear scope of applicability.
Back to career directionsMathematical and Computational Methods Researcher
Abstract structure from scientific problems and develop mathematical, inference, optimization and numerical methods.
Example contribution: Mathematical and Computational Methods Researcher
A theoretical result, algorithm or numerical method with explicit assumptions, accuracy and conditions of use.
Back to career directionsResearch Software Developer
Turn scientific methods into correct, clear, maintainable software that others can reuse.
Example contribution: Research Software Developer
A research software package with appropriate validation, documentation, examples and version history.
Back to career directions
Students can focus on one or two roles while understanding how related parts of research connect. We agree on development goals together, taking account of each student’s stage, interests and the support they need.
These examples illustrate possible contributions; individual goals are agreed together.
Where these capabilities can lead
We explicitly encourage students to explore these career directions. Job titles and responsibilities vary across institutions, and some paths overlap. Students can choose, combine and revise their goals over time.
Use these connections to choose research tasks and experiences. The capabilities are possible starting points; each student can choose a focus, combine strengths and revise their direction over time.
Professor or Faculty Member
Conduct research, teach and mentor students, and develop an academic direction.
Explore this path: Professor or Faculty Member
- Capabilities to explore
- A possible contribution
- Complete a study that connects a question, methods and evidence, and develop an explanation that can be used in teaching.
- Experience to build
- Develop depth in a method or scientific domain, and build experience in teaching, mentoring and academic writing.
Principal Investigator
Set a research agenda, organize a team and collaborations, and take responsibility for research and training.
Explore this path: Principal Investigator
- Capabilities to explore
- A possible contribution
- A research agenda with connected questions, an initial study and a feasible plan for collaboration and follow-up work.
- Experience to build
- Build a deep research specialty and gain experience in proposal development, mentoring, collaboration and team and resource management.
Research Scientist
Pursue sustained research on scientific questions, methods and discoveries.
Explore this path: Research Scientist
- Capabilities to explore
- A possible contribution
- A scientific study with reproducible methods, a traceable chain of evidence and clearly stated limits to its conclusions.
- Experience to build
- Build sustained research practice and domain knowledge, with depth in mechanistic research or mathematical and computational methods.
Machine Learning Research Scientist
Investigate learning methods, scientific representations, model architectures and generalization.
Explore this path: Machine Learning Research Scientist
- Capabilities to explore
- A possible contribution
- Propose a method and use controlled comparisons and tests across settings to establish its effectiveness and limits of applicability.
- Experience to build
- Deepen mathematical, statistical and machine learning foundations while developing a distinctive methodological contribution.
Applied Scientist
Translate research methods into verifiable improvements in industrial R&D or products.
Explore this path: Applied Scientist
- Capabilities to explore
- A possible contribution
- An applied prototype that demonstrates a measurable improvement on a real R&D or product task under relevant operating constraints.
- Experience to build
- Learn the application domain, user needs and operating constraints through collaboration and practical projects.
Machine Learning Engineer
Implement training, inference and model integration, and verify operational performance.
Explore this path: Machine Learning Engineer
- Capabilities to explore
- A possible contribution
- A reproducible training and inference pipeline with measured model quality, runtime performance and documented update procedures.
- Experience to build
- Build software engineering and deployment experience, including monitoring and performance troubleshooting.
AI Systems Architect
Coordinate system components, interfaces, feedback, performance and reliability.
Explore this path: AI Systems Architect
- Capabilities to explore
- A possible contribution
- Build a system prototype with explicit interfaces, feedback and resource budgets, and test overall performance and failure recovery.
- Experience to build
- Build experience across multiple system components and learn to make and explain architectural trade-offs.
Laboratory Automation Engineer
Connect instruments, software, robotics and experimental workflows.
Explore this path: Laboratory Automation Engineer
- Capabilities to explore
- A possible contribution
- An instrument-and-software workflow that connects experiment execution, data capture and the next experimental decision with traceable records.
- Experience to build
- Gain supervised experience with instruments, control software and robotics where relevant.
R&D Lead
Organize technical direction, team collaboration and R&D outcomes.
Explore this path: R&D Lead
- Capabilities to explore
- A possible contribution
- A coordinated R&D project with a clear technical direction, connected contributions and evidence that supports key delivery decisions.
- Experience to build
- Develop technical depth and gain experience in team coordination, mentoring and resource planning.
Scientific Entrepreneur or Technical Cofounder
Turn technical discoveries into products or services and build a team for sustained development.
Explore this path: Scientific Entrepreneur or Technical Cofounder
- Capabilities to explore
- A possible contribution
- A technical prototype with evidence of a real user need, documented feedback and a practical plan for a pilot product or service.
- Experience to build
- Build technical depth and gain experience in customer discovery, product development, team building and business operations.
Explore more career directions
Scientific and domain research (5)
Computational Materials Scientist
Study relationships among material structures, properties, conditions and responses.
Explore this path: Computational Materials Scientist
- Capabilities to explore
- A possible contribution
- A computational study that connects material structure and conditions to a target property and tests a proposed mechanistic explanation.
- Experience to build
- Build materials knowledge and experience with suitable simulation methods and experimental interpretation.
Computational Biologist
Model biological systems and perturbations to develop testable biological explanations.
Explore this path: Computational Biologist
- Capabilities to explore
- A possible contribution
- A model of a biological system whose predictions under perturbation help distinguish competing biological explanations.
- Experience to build
- Build biological knowledge and experience interpreting experimental designs, measurements and uncertainty.
Bioinformatics Scientist
Analyze sequence, omics and other data to answer specific biological questions.
Explore this path: Bioinformatics Scientist
- Capabilities to explore
- A possible contribution
- A reproducible sequence or omics analysis that answers a biological question and checks statistical assumptions and data quality.
- Experience to build
- Build relevant biological and statistical knowledge and experience with the data types used in the field.
Molecular Design Scientist
Design and validate candidate molecules with desired properties or functions.
Explore this path: Molecular Design Scientist
- Capabilities to explore
- A possible contribution
- A candidate design study with explicit objectives, a feasible molecular search space and evidence that tests the proposed function.
- Experience to build
- Build knowledge of the molecular domain and learn how candidate designs are assessed experimentally.
Experimental R&D Scientist
Design and conduct experiments, connecting measurements, mechanistic judgment and R&D decisions.
Explore this path: Experimental R&D Scientist
- Capabilities to explore
- A possible contribution
- A controlled experimental study that links measurements to a mechanistic judgment and a justified next R&D decision.
- Experience to build
- Gain supervised laboratory practice with the relevant instruments, measurements and experimental procedures.
Scientific computing, software and engineering (5)
Scientific Computing Researcher
Develop numerical methods, simulations and approaches to inverse problems.
Explore this path: Scientific Computing Researcher
- Capabilities to explore
- A possible contribution
- A numerical method or inverse-problem solver with accuracy and stability comparisons, documented assumptions and reproducible examples.
- Experience to build
- Deepen numerical analysis and scientific computing practice, including implementation and verification.
Research Engineer
Build research prototypes that turn complex ideas into testable experiments.
Explore this path: Research Engineer
- Capabilities to explore
- A possible contribution
- A working research prototype that makes a complex idea testable and supports clear comparisons with alternative designs.
- Experience to build
- Practice translating research ideas into reliable implementations and collaborating closely with researchers.
Research Software Engineer
Develop, validate, maintain and extend research software.
Explore this path: Research Software Engineer
- Capabilities to explore
- A possible contribution
- A reusable research software package with appropriate validation, documentation, examples and a maintenance history.
- Experience to build
- Build software design, testing, collaboration and maintenance experience alongside an understanding of the science.
Scientific Data Engineer
Build processes for data collection, processing, quality checks and traceability.
Explore this path: Scientific Data Engineer
- Capabilities to explore
- A possible contribution
- A traceable scientific data pipeline with documented transformations, quality checks and versioned outputs.
- Experience to build
- Learn the scientific meaning of the data and gain experience with databases, data models and reliable processing.
ML Platform Engineer
Build shared platforms for training, computing, experiment tracking and reproducibility.
Explore this path: ML Platform Engineer
- Capabilities to explore
- A possible contribution
- A shared training and experiment-tracking platform that another research project can use and reproduce from its documentation.
- Experience to build
- Gain experience with computing infrastructure, operational reliability and supporting research users.
Evaluation and product development (2)
AI Evaluation Researcher
Design evaluation tasks and tests, and study model capabilities and failure boundaries.
Explore this path: AI Evaluation Researcher
- Capabilities to explore
- A possible contribution
- An evaluation protocol that tests a clearly defined capability, detects shortcuts and identifies conditions under which the model fails.
- Experience to build
- Develop experimental design, statistics and domain expertise for interpreting model behavior.
Scientific AI Product Manager
Translate researchers’ needs into product goals, features and acceptance criteria.
Explore this path: Scientific AI Product Manager
- Capabilities to explore
- A possible contribution
- A product brief and prototype that connect a research user’s needs to measurable goals and evidence from a small user trial.
- Experience to build
- Gain experience in user research, product prioritization and collaboration with research and engineering teams.
We support members in exploring these paths and developing the specialist skills they need through courses, co-mentoring, internships and further training. Experimental R&D, laboratory automation, product development and entrepreneurship also require relevant practical experience and working conditions.
03 · Shared commitments
Guidance, autonomy, and responsibility grow together.
The PI provides research direction, shares reasoning, creates development opportunities, and matches responsibilities with guidance and resources. Project Leaders connect daily work to the scientific question and respond to difficulties. Students bring their own thinking, communicate honestly, and follow through on agreed work.
Commitments include the PI's reviews and decisions. When work is delayed, we examine the task, workload, guidance, and support before deciding how to adjust. Members may question anyone's scientific views, including the PI's, and should be able to seek help and report errors without retaliation.
Read the corresponding charter section04 · Daily practice
Deep engagement, sustainable effort, and room to explore.
We protect uninterrupted time for reading, reasoning, experiments, and writing. Shared working arrangements are coordinated by project. There is no lab-wide clock-in requirement or routine requirement for evening or weekend overtime. Worthwhile learning and exploration can be valuable before they produce a paper.
We recognize reliable research, useful code and tools, mentoring, and contributions to shared resources. We explicitly encourage academic, research, AI engineering, laboratory automation, R&D leadership and technology entrepreneurship paths, while supporting other directions and regular conversations about growth. The full charter explains our arrangements for feedback, research integrity, authorship, recognition, working relationships, and graduation or departure.
Read the corresponding charter section