Abhishek Sharma
⚠️ I am looking for industry positions. Let's talk! ⚠️
SEAS, Harvard University
Boston, MA
I am a fifth year Ph.D. student in the Data to Actionable Knowledge (DtAK) lab at Harvard University, where I am advised by Finale Doshi-Velez.
My research focuses on reinforcement learning (RL) and representation learning for decision-making. I create models that are not only predictive but also align with important properties, such as interpretability and safety.
I have collaborated with hospitals (Massachusetts General Hospital, Brigham and Women’s Hospital) and the Singapore government (Temasek Laboratories) on applying machine learning to healthcare challenges. During my internship at Google, I worked on developing interpretable foundational models for waveform data. My PhD projects include creating interpretable and safe batch RL methods to improve clinician decisions. I have also applied my work to depression subtype discovery and disease progression modeling.
Before Harvard, I earned a M.S. in Computer Science from the University of Massachusetts, Amherst, where I worked with Madalina Fiterau and Philip Thomas. Even prior to that, I obtained a bachelor’s and a master’s degree in Materials Science and Engineering at IIT Madras, India, where I worked with Nandan Sudarsanam.
news
Nov 20, 2024 | New preprint on arXiv: Inverse Transition Learning: Learning Dynamics from Demonstrations to learn dynamics from expert demonstrations while preserving the optimality of the expert policy. |
---|---|
Oct 11, 2024 | New preprint on arXiv: Decision Points RL (DPRL) to identify “diffs” to the behavior policy in batch RL settings. We achieve provably high-confidence improvement. |
Sep 01, 2024 | Finished my internship at Google Research (more details and paper soon!). |
Aug 01, 2024 | Organizing the RLC 2024 ICBINB workshop. The workshop will celebrate innovative RL research that led to counterintuitive results. |
Jun 03, 2024 | Started my internship at Google Research with Farhad Hormozdiari and Justin Cosentino. I will be focusing on foundational modeling efforts on waveform data! |