As background context, I am aggressively protective of my time, as I believe we all should be.
On a typical day, I receive over a hundred emails. Please do not be offended if I do not reply to yours
or redirect you to read this page. If I answered every email I receive, I would not be able to do much else
(and if all I did was answer emails, why would you want to talk to me?).
As a professor and a researcher, I wear both hats -- of a maker and a manager. I encourage you to read
this excellent essay about why disruptions
are so corrosive and disruptive to the maker's schedule.
Here are answers to the questions I am frequently asked over email.
By reading these, you are helping me cut down on the time spent answering email (and actually get work done!). I already like you.
Prospective Students, Postdocs
I want to join the MS Program @ IC/CS/CSE GT.
Please see the admissions page.
I play no role in this process. Please do not email me.
I want to join the CS/ML/Robotics Ph.D. Program @ GT and work in your lab.
Great, happy to hear of your interest.
I am always looking for sharp motivated Ph.D. students with interests in machine learning, computer vision, NLP, and AI.
Yes, all PhD offers made by IC GT are fully funded.
In your application, please pick the "Computer Vision", "Machine Learning", or "Computational Perception" sub-areas.
In your application, please mention my name so I can easily find it.
Note: admission decisions are made by a committee and not by individual faculty. I cannot reply to emails asking me to assess admission chances.
I am looking for a post-doc position.
Great, happy to hear of your interest. I am currently looking for a postdoctoral researcher with a
strong background in machine learning, computer vision, or natural language processing.
Please send me an email with a CV and a link to your top 2-3 papers in the field.
What do you look for in a student?
[Partly inspired by Prof. Devi's Parikh's note]
At a high-level, here's what I look for in a student, or for that matter any collaborator (in roughly decreasing order of importance):
Clarity of thought (e.g. why are you doing what you're doing?) and logical thinking (e.g.
do non-sequiturs in arguments bother you?)
Clarity of expression (e.g. can you explain things to me without jargon or incoherence?)
and theory of mind (do you understand what your interlocutor knows/believes/understands?)
Attention to detail and ownership (e.g. do you take pride in what you produce? Does the idea of
crap with your name on it deeply disturb you?)
Reliability (e.g. do you do what you say you'll do?)
Ability and desire to pick up new skills (e.g. how long does it take you to pick up a new 'framework'?)
Proficiency in skill-set generally required in technical fields (math, statistics, programming).
Notice that most people tend to focus on the last bullet. It's important but (according to me) not the most important.
I am looking for a visiting research position / internship (typically 6 months - 1 year).
If you are not already at GT, I am unable to host you during the 2021-22 academic year.
I am leaving the instruction below because they may have helpful pointers.
Great, happy to hear of your interest.
First, some context:
Yes, my group has in past and does frequently host one year interns/visitors.
Please see this list to see the profiles of people we have hosted in past, and where they are now.
Typical duration is 1 year. Anything shorter than 6 months isn't productive. No summer internships.
I believe we are outliers in the sense that most academic groups do not generally have such positions.
These positions are similar in nature to the residency programs that are now popular in industry (e.g. FB, Google, NVIDIA residencies).
Why do I do it? Because people gave me a shot and it changed my trajectory. I am paying it forward.
Having said that, the bar for entry has become very high over the years. I receive a large number of applications and can host a small number.
As described here, my lab has fairly broad and
interdisciplinary interests (vision, NLP, machine learning, AI). Projects in my lab span the theory-to-applied spectrum.
This and this are examples of more theoretical projects.
This is an example of a more applied project. And this
is somewhere in between.
Here's what the process looks like (partly inspired by Devi's note):
You send me an email with the following information :
Subject: [position] [time], where [position] = intern, [time] = Fall 20xx, May-December 20xx, etc
Link to your webpage and CV
2-3 sentence description of your past projects (highlighting ones that you believe prepare you for research in my lab)
2-3 sentence description of what you are interested in working on (it's okay to be non-specific here, we'll likely iterate on this if you're selected)
Link to your github page
Most important bullet: Link to one of your code repositories that you believe
is your strongest accomplishment (in terms of software engineering, research, or a mix). Maybe you reimplemented/replicated a published paper,
or maybe you made an existing library 10x faster/lighter/smaller/better, or maybe you successfully landed a sophisticated pull request
into a large open-source codebase (like OpenCV, CloudCV, etc). Basically, please show me something you are most proud of (and please tell me why).
For calibration, here are repositories I believe their creators should be proud of:
example 1example 2,
(notice the well-structured well-documented repositories, systematic progress, good software
engineering principles; this is the kind of code you should be producing).
What do you believe is your biggest strength? What percentile would you place yourself at amongst your peers in this regard?
What do you believe is your biggest weakness? What percentile would you place yourself at amongst your peers in this regard?
In what way have you improved the most in the last year? How?
Along what weakness have you not improved but you think that’s fine / is low priority, and why?
Georgia Tech Students
I want to work with you.
Great, happy to hear of your interest. I hope we are able to work together. A few caveats:
The best way to interact with me is via my group.
You are welcome to email my group to find out what they work on,
and how you may contribute. Having read their published papers will help to start a dialog with them.
They are often looking for collaborators and junior students to mentor.
Please understand that they may be busy and may not always be able to respond.
Unless you already have significant experience in Machine Learning, Computer Vision, or AI, I will ask you to take a class
with me first
(e.g. my Deep Learning class),
and work on a solid class project. The best class projects are typically the starting points for longer-term research engagements.
Please read the answers above about what I look for in students. You are welcome to follow the instructions to apply as an intern
in the lab. Please note: unless we are already working together or you have already found someone in my group to mentor you,
I generally do not take on new "special problem" project supervision.
As a general principle, I do not fund students until we have had a chance to interact and work together on something (maybe a class project).
I would advise against sending me cold emails looking for funding.
Can you serve on my committee?
I am generally honored to serve. One caveat -- if your thesis or research topic is related to Machine Learning and/or Computer Vision,
it would help if you have already taken
relevant courses (see above) before you approach me (and no, watching lectures online, even if they are mine, don't count the same as taking the class).
I find without these basic classes, we often don't have a language to communicate in, and you will not benefit from my involvement.
Press and Commercial Inquiries
Can you comment on a story?
In general, I'm happy to talk -- either about research from my group at Georgia Tech or comment on work from peers.
If the research was conducted by Facebook AI Research (FAIR), I will ask you to redirect your questions to email@example.com.
Can you help my company / startup with a Machine Learning project?
I receive a number of such requests.
Please note that funding the research work of a typical Georgia Tech ML PhD student costs between $80k and $100k per year.
Some students may be open to consulting, but we generally do not give away our time for free. If you are interested in such opportunities,
please get in touch.