The Machine Learning Center at Georgia Tech (ML@GT) is home to many talented students from across campus, representing all six of Georgia Tech’s colleges and the Georgia Tech Research Institute (GTRI).
These students have diverse backgrounds and a wide variety of interests both inside and outside of the classroom. Today, we’d like you to meet Samarth Brahmbhatt, a recent graduate from ML@GT whose work focuses on teaching robots to grasp objects as well as humans can.
Name: Samarth Brahmbhatt
Hometown: Gandhinagar in the state of Gujarat, India
Advisor: James Hays, School of Interactive Computing. I also work closely with Charlie Kemp in Biomedical Engineering.
Current Georgia Tech degree program/year: Ph.D. Robotics, Fall 2014 – Spring 2020
Other degrees earned and from what institution: MS in Robotics from the GRASP Lab at the University of Pennsylvania and B. Tech. in Electronics and Communication Engineering from Nirma University
Tell us about your research interests. Where might people be impacted them in everyday life?
My Ph.D. thesis has focused on hand-object contact during grasping. This kind of contact has been approximated with varying degrees of accuracy in decades of previous research in computer vision and robotics.
What makes my work unique is that we have developed methods to directly observe this contact with high resolution thermal cameras. This has enabled us to confirm some intuitions about our grasping behavior, and has also revealed some new insights. For example, we tend to use only the tips of our thumb, index, and middle fingers to grasp objects for handing them off, but use other fingers and areas of the hand to grasp objects for using them. We have also used this data to learn hand-object contact models which are more accurate than previous heuristic-based methods. These contact models are basic ingredients of algorithms in robotic grasping and hand-object interactions in virtual reality (VR). This work also advances our scientific understanding of our grasping ability.
More broadly, I am interested in developing robots that can help us in our homes and hospitals. Being able to recognize, pick up, and use everyday objects is a big part of that. My PhD work focused on observing how we humans do it, and I’m planning to use lessons learnt from that for robotics work in the future.
What drew you to wanting to research these areas?
For me, it was much more of a progression of ideas rather than a childhood dream. I realized in college that working at the intersection of hardware and software made me happy, and researching new ideas made me happy. That drew me to a master’s degree in robotics, where I realized that grasping is a really challenging problem for robots. But we humans do it so well! So when the idea of studying human grasping behavior came up in a brainstorming meeting with my Ph.D. advisors, I got excited about pursuing it further.
What first drew you to machine learning?
You cannot escape machine learning if you are working in computer vision or robotics 🙂
What inspired or motivated you to pursue your Ph.D.?
I worked with Ph.D. students on a couple of research projects during my master’s, and interacted closely with them. That gave me a close-up view of what a Ph.D. is all about. It was very appealing to me that you get paid to create and disseminate new knowledge. On top of that, you get to learn — from some of the best professors — how to think critically, make evidence-based arguments, and share your discoveries in compelling papers and talks.
Earning your Ph.D. requires a lot of work. What has been challenging, rewarding, or unexpected about this experience?
It was challenging to realize that a lot more than academic research goes into finishing a Ph.D. Health (especially mental health), finances, social relationships inside and outside the lab, and a lot of other things enter the picture.
What’s a good piece of advice you have received?
Advice from some mentors about defining a good Ph.D. project has been very helpful. A good Ph.D. project generates new knowledge, uses your strengths, and avoids jumping into crowded research areas where your contributions are likely to be only incremental.
What have you been up to during quarantine?
The quarantine has been pretty busy for me because I defended my Ph.D. thesis at the end of April. Since then, I have been busy with wrapping up research loose ends and getting ready to move to California.
What’s the most interesting or random app on your phone?
What’s in your fridge right now?
Lots of Theo dark chocolate with almonds.
We’re at the beginning of a new decade. What are you most looking forward to in the next ten years?
I am looking forward to more cooperation and increased efforts to address global warming.
What place is at the top of your travel bucket list and why?
I have always wanted to try a long ocean trip by ship to see the sky full of stars and learn how ancient mariners navigated.
Who is someone that inspires you and why?
My parents, because of the significant odds they have overcome in life.
What is your proudest accomplishment so far?
Managing to learn swimming flip turns from YouTube videos!
Podcast, movie, tv show, or book? Why that medium and what are some of your favorites from your chosen medium?
Podcast, definitely. Books have a special place in my heart, but lately I’ve found that podcasts are much more conducive to my lifestyle. I like the EasyGerman podcast — it is great for getting German practice, and they discuss interesting topics!
Why do you think embodying Georgia Tech’s motto of “progress and service” is important, especially in regards to ML and AI?
While progress in these powerful technologies is obviously desirable, it is equally important to keep in mind how they can serve humanity.
You’ve just graduated from Georgia Tech. What’s next?
I’m moving to California to work as a post-doc with Vladlen Koltun in the Intel Intelligent Systems Lab. I can’t wait!