My mentoring program

Thu, 16 Apr 2026 00:00:00 +0000

Intro

Coming where undergrads are heavily involved in research, I was surprised to see so little upon arrival at Georgia Tech. I was told it was hard to get them up to speed to the point where they would be productive, and that one needed a very well-defined project in order for them not to get lost. Basically, that it was unlikely to accelerate the pace of research overall. Nevertheless, it seemed to me that there was an unexploited opportunity to provide meaningful experiences to willing students without sacrificing the pace of my own research.

Once I gained confidence in my research path and had some projects that were sufficiently de-risked, I began recruiting students. I started with one, then later 4, and kept expanding to up to 25 students at a time as I saw that the research output we were producing was more than I would have achieved with the time I put in. Most of the time cost is at the beginning: defining projects, recruiting, and setting things up. The keys to all this?

Minimal handholding—I like to think of it as throwing them into the deep-end, but with life preservers (in the form of clearly laid-out preparatory material and to-do lists).
Quantity—Student output has been widely variable and hard to predict, with some producing little more than a few lines of code in a year and others implementing a whole section in a paper or even leading a team of other students as first authors. Once I learned this, I began over-recruiting, hedging my bets, and letting the cream rise to the top. Another benefit of this approach is that mentees can often help each other, rather than turning to my as the first recourse.

This might not work everywhere, but at GT we have the luxury of an abundance of bright students seeking research opportunities, so we don’t need to be afraid of scaring off those who aren’t as well prepared or self-reliant. Extended to the internships and entry-level jobs, this may indeed be the state of the world economy in general, as STEM education seems to have outpaced STEM entrepreneurship/job creation in many fields and places.

What follows is the recipe for the program I’ve developed, written in a mix of a prescriptive and descriptive style, as I’ve tried to simultaneously

minimize risks to my time, research output, and path to graduation, and
maximize the likelihood of success, both for the mentee and for me.

Defining the project

In the spirit of minimizing risk, the project should be either

a “nice to have,” “side quest” sort of project, or
an “I don’t need this yet” sort of project. This could be a limited “scout ahead” or “pave the way” assignment or, if you have capable-enough mentees, a whole paper or piece of software.

Otherwise, if you have someone working on something you actually need soon for your dissertation and they don’t get it done in time, you’ll have to step on their toes to finish the job.

The ideal project is also one that is well-defined in your mind, with clear methods and deliverables. It helps to create a roadmap for the project so the mentee will always know what they’re supposed to be working on. Otherwise, your mentee will likely depend on you figuring things out as you go, which will be a drag on their time and yours. Most students will not have the background to be productive initially, so it helps to include a training phase with learning material at the beginning.

Example roadmap

All-optical control project roadmap

Learning material
- Basics of simulating spiking neurons: Chs. 1-3 of Neuronal Dynamics
- Cleo paper
- Brian tutorials/user guide
- Cleo tutorials
- Good Research Code ebook
- Collaborating section of Atlassian git tutorials
Project
- Holographic stimulation
  - Write tests
  - Define 2D plane, select neurons in focus
  - Gaussian ellipsoid light propgataion model
  - Find out from experts if we need to model raster scanning constraints: update frequency? periodic rather than constant stimulation?
- 2P imaging
  - Write tests
  - Get distance from focal plane
  - Define generic indicator interface: from voltage? From spikes?
  - Implement calcium indicator (equations from NAOMI)
  - Output trace for each neuron with appropriate noise
  - Find out from experts: Cross-talk from light source activating indicators? Max sampling frequency?
- Create a tutorial implementing all-optical control for inclusion in docs

Only consider supervising an open-ended project (where you don’t know exactly how you’d do it) if you’re sure the student is highly independent and capable, i.e., if you’ve already worked with them and seen impressive results.

Preparing recruiting materials

You’ll need a project description for potential recruits. Things you will probably want to include:

Background/context for the problem
What the student will be doing—you may want to use the roadmap you prepared here
Skills required, desired, and/or to be learned on the job
Will research be for credit? Pay? Volunteer?
- I’ve found that students are much more consistent when working for credit than on a voluntary basis. It really makes a difference when a grade depends on working X hours a week.
Expected time commitment: hours per week, number of semesters
Will work be in-person, remote, or hybrid?
Will the work be solo or as a team?
- A surprising number of students turn out to be not very productive out of capability or just time constraints, so if you assign multiple students to a project that only needed one stellar student, the probability of success rises significantly. Many are called, but few are chosen. Just make sure that if you do so, each person has well-defined, non-overlapping responsibilities so they don’t disengage out of boredom or the sense that someone else is already working on it. And if it’s too hard to break up a project in this way, another solution is to switch people to another project—another benefit of running large program with multiple projects.
How to apply
- Since other grad students in the were also interested in mentoring and it was a pain to coordinate when interested students emailed, we developed a shared research interest form with their interests, background, availability, etc.: . This way all potential mentors could see applicants and we could mark on the database who had contacted who.

You may want to prepare a process to screen applicants before you interview them—for example, a technical challenge involving skills they’ll need for the project. The project is much, much more likely to be successful if you pick the most capable students. At a minimum, you will filter out applicants who are only mildly interested and get a better measure of skill level than you can guess from their résumé. Occasionally, this can reveal truly stellar students (for example, an incoming freshman who can implement spiking neuron models, optimal control, and neural networks!). : I asked them to fill out different sections, depending on which project they were interested in. My lab-mate was the inspiration for this with a he made.

Advertising

With your project description done, you are ready to advertise. Here are some things that worked for me and others at Georgia Tech:

PIs (starting with my advisor) — many get emails from proactive students looking for research opportunities
A campus-wide mentored position aggregator, like Georgia Tech’s
A centralized office for coordinating research like GT’s , if one exists
Program-specific channels — email lists, academic advisors, clubs, etc.
Campus-specific social media (Reddit, Discord, etc.)

You will probably want to advertise at least a week before classes start and likely earlier, so students have time to plan research into their schedule.

Interviewing

You will want to define the criteria you’re looking for ahead of time and take notes. For example, I would rate students’ skills in areas relevant to the projects, as well as social/professional skills, time commitment, etc. Unless you’re really sure, you should probably avoid giving a sure “yes” or “no” in the interview itself. That way, you can think more objectively and reflect without the social pressure of the moment. Another approach, which I adopted later on, was to accept pretty much everyone, since there was almost always some project they could accelerate after getting up to the speed if they had the talent and dedication—many hands make light work.

Consider interviewing more students than you think necessary; a surprising number of students drop out because of other opportunities or because their schedule changes.

Onboarding and administrative tasks

This will of course highly depend on the resources at your institution, but it helps to write down the repeated steps of the onboarding process and keep track of what has been done and what hasn’t. For example, Slack access, building access, sending them their team/project assignment and roadmap, signing up for credit if applicable, etc. I found I had rules, expectations, and advice I wanted to transmit to all mentees, so I prepared a “guide” document with the following sorts of things:

how often you have formal meetings and how you expect the mentee to prepare for them
your availability for troubleshooting: time limitations, office hours, etc.
how much time you expect them to spend in the lab in person, if their work can be done remotely
wisdom or principles to maximize their chance of success
recommended software infrastructure

Some other administrative miscellany:

To regularize grading, the lab developed a uniform syllabus. Well, two: and the other for data collection (for helping run human subjects experiments).
It helps to centralize information regarding course numbers and department-specific instructions for signing up for research credit, since those questions come up again and again.
We deep a Notion database with all mentees, including name, supervisor, project(s), program, year, major, what lab resources they have access to, etc.

Working procedures

I would hold weekly update meetings where all mentees present a short 1-2-slide update. Later as some teams got to be large, I split into a few different meetings. I would often remind mentees to reach out if they are stuck and needed a troubleshooting session, but one-on-one meetings were not the default.

Expansion

If you have enough well-defined projects for multiple students, why not mentor multiple at the same time? If you structure things right, scaling up can increase yield on your time, since you can meet multiple students at once and (if the project areas or tools overlap) the mentees can help and teach each other. Here are some of the main things I’ve tried to foster subgroup collaboration:

Have students ask troubleshooting questions on a common Slack channel so they can help each other or at least see your answers for other people
In the weekly update meeting, have people include background/context for their project in their report so others on different projects can follow what’s going on
Encourage questions and input from other mentees when someone is reporting on their progress

Another way to expand is by adding yet another level to the hierarchy: once I had an abundance of mentees and recognized a few who excelled and were capable of working quite independently, I made them leaders of team projects. This accelerated research further without my time constraints serving as a bottleneck.

End-of-semester report

Students researching for credit or pay are required to turn in an end-of-semester report, and volunteers should do so anyway. It may help them to answer the question, “What would someone else need to know if I disappeared and they had to pick up where I left off?” This could include:

a recap of the problem
different tools, methods, or experiments tried and their results
where to find/how to navigate relevant code and data
current state of the project, next steps

Semester recap & reflection meetings

It is useful at the end of the semester to meet individually with each trainee to look back at how things went and discuss future plans. You may want to cover topics such as:

What went well? What didn’t?
What did you (the mentor) do well? How can you improve? How could the mentorship processes and structure be improved?
What did they (the mentee) do well? What would they have done differently? What did they learn?
What are their future plans? Do they want to continue research in the lab? On the same or on a different project?
If they did outstanding work and you think they’d merit a paid position, do they want to continue for credit or for pay?

The feedback I received from these meetings helped me refine the program from semester to semester.

Conclusion

Again, in my experience, the initial investment in setting this all up and the week-to-week mentoring time has more than paid for itself in terms of research output. At the time of writing, ~14 total mentees are on track to be coauthors on papers, and three papers executed mostly by mentees are on track to be completed. I’d like to think this has been beneficial to my mentees as well, many of whom have gone on to internships and a few have gone on to grad or medical school. Hopefully readers find my experience useful, whether for academic research or industry internship settings, where I assume many principles would carry over. Good luck!

Mentoring |