Overview

I am building a research group at BYU and recruiting graduate students to it. The work centers on sequential decision making where the usual assumptions fail — partial observability, delayed feedback, confounded data, and actions that cannot be taken back.

I advise students to become researchers who can pick their own problems, not assistants who execute mine. That means handing over questions early and pushing you to defend your own choices about what is worth pursuing.

My job is to help you think concretely about the problem in front of you: to say precisely what you are claiming, what would show it false, and what you would have to measure to know either way. That habit is what I am actually teaching.

The other half is getting you to enjoy the process rather than only the results. Most of research is things not working, and a career built on external validation makes those stretches miserable. I would rather you chase a question you genuinely want the answer to. Publications follow from that more reliably than the reverse. I wrote about this at more length in Returning to Academia.


What we work on

  • Offline reinforcement learning — learning to act from fixed datasets, including when the data records only what someone already chose to do
  • Irreversibility and risk — identifying states from which good outcomes are no longer reachable, and acting early enough to matter
  • Delayed and unreliable feedback — credit assignment when outcomes arrive late, or when the signal you train on is a fallible stand-in
  • Structured action spaces — decisions composed of many interacting parts, where enumeration is not an option
  • Clinical and scientific decision making — the applied settings that generate these problems in the first place

See the research overviews for the specific themes and the prior work behind them.


How I advise

Weekly one-on-one meetings. Your project, your agenda. I come prepared to argue with you about it.

Problem selection is taught, not assigned. Early on I will hand you a scoped question. Over time you take on more of the choosing, because deciding what to work on is the skill that outlasts any particular technique.

Your development is the point. The measure I care about is how much stronger a researcher you are when you leave than when you arrived — not how much my publication list grew while you were here. When those two pull in different directions, I will side with the first.

Write early and often. Not just at submission time. Writing is how you find out whether you understand what you are claiming.

Negative results get reported. If an experiment kills a hypothesis, that is progress and we treat it as such. I would rather find out in week three than month six.

Reproducibility from the start. Organized code, tracked experiments, results someone else can regenerate. This is not bureaucracy; it is what makes the third year of a PhD survivable.

Explicit career conversations. Academia and industry are both legitimate destinations, and I have worked in both. I will help you prepare for the one you actually want rather than the one that reflects well on me.

Sustainable pace. I have worked at the frontier under permanent deadline pressure and know what it costs. Ambitious research does not require running yourself into the ground, and I will not ask you to.


What I look for

  • Genuine interest in decision making under uncertainty, and some evidence of it
  • Strong programming ability — you will implement things, not just describe them
  • Mathematical maturity in probability, linear algebra, and optimization
  • Willingness to work on problems where the right formulation is not yet known
  • Directness. I would rather hear that an approach is not working than discover it later.
  • Interest in the reasoning itself, not just the result — telling a correct argument from a merely fluent one is increasingly the scarce skill.

Prior publications are not a requirement. Curiosity and persistence matter more, and I can teach the rest.


Prospective students

Applying to BYU CS. Mention my name and your interests in your application, and email me separately. I read what students send me.

Already at BYU. Come talk to me. Reading-group participation and small projects are the usual entry points, and both are open.

What to send. A short email covering your background, one or two problems that interest you and why, and what you are hoping to get out of a PhD. Concrete beats polished — a paragraph about a specific question tells me far more than a list of coursework.

Contact: tkillian@cs.byu.edu


I also mentor undergraduate researchers, and several projects are structured so that undergraduate and graduate students work on adjacent pieces of the same question.