Siddartha Devic
devic [at] usc.edu
Computer Science Department
University of Southern California
About
I’m Sid, a CS PhD student in the USC Theory Group, where I am fortunate to be advised by the wonderful Vatsal Sharan and Aleksandra Korolova. I am broadly interested in trustworthy, fair, and robust machine learning and AI. I am grateful to have my research supported by a National Defense Science and Engineering Graduate (NDSEG, 2021-2024) Fellowship and a USC Viterbi Graduate Fellowship (2024-2025). Here is my CV (Sep. 2025).
I have been lucky to spend some time doing fun research at various places outside of USC during my PhD.
- Summer 2025: Research internship at Apple with Aravind Gollakota and Preetum Nakkiran.
- Fall 2024: Visitor at the Simons institute modern paradigms in generalization research program.
- Summer 2024: Research internship at Amazon with Gaurush Hiranandani.
I completed my undergraduate degrees in math and CS at The University of Texas at Dallas (UTD), where I explored research in networking, theory, optimization, and reinforcement learning. Along the way, I was extremely fortunate to be advised by these amazing professors, who greatly shaped the way I think about problems: Nick Ruozzi, Jason Jue, Brendan Juba, and Ben Raichel.
If you are a prospective PhD applicant, I am happy to help in any way I can. Please feel free to reach out over email.Publications and Preprints
(*) indicates equal contribution.
Uncertainty Quantification and Calibration
Machine-learned models and systems which "know what they don't know" is a critical requirement of trustworthy AI. How should we measure and quantify such model uncertainty in practice? How does the problem change when models are deployed in the wild, especially in settings with diverse and non-homogenous populations? Finally, what does uncertainty quantification mean for large language models, and how is it impacted by additional training?
Trustworthy Algorithmic Marketplaces and Recommender Systems
Modern algorithmic platforms like LinkedIn, ride-sharing, or content delivery are increasingly complex aggregations of multiple machine-learned systems. How can we ensure that these "composite" systems are trustworthy, fair, and reliable? How does the problem change when pieces of these systems have intrinsic uncertainty tied to their predictions?
Foundations of Machine Learning
Our work in this area explores some basic questions in (efficient) learnability, the power of unlabeled data, and the role of regularization. A lot of this work was motivated by a beautiful object called the one-inclusion graph, which reframes supervised learning as a bipartite matching problem. Our collaborator Shaddin Dughmi wrote a wonderful expository article overviewing this perspective and some of our work.
Other
Additional research spanning networking, reinforcement learning, and black-box optimization.
Mentorship
-
Alex Reyes Aranda and Kuan Liu - Phenomenological deep learning.
Undergrads, May 2025 - Present. -
Nathan Derhake and Kuan Liu - Auditing for multicalibration.
Undergrads, Oct. 2024 - Present. Part of USC CURVE program. -
Dutch Hansen - Multicalibration of neural networks.
Undergrad, Jan. 2023 - May 2025. Part of USC CURVE program. Dutch is starting his PhD @ UW in Fall 2025. Look out for him! -
Anish Jayant - Multicalibration of neural networks.
Undergrad, Jan. 2023 - May 2023. -
Jayron Martinez - Bias in Machine Learning Algorithms.
High school student, summer 2022. Part of the USC SHINE program. Jayron started his BS @ USC in Fall 2024. Look out for him!
[poster]
Service and Teaching
Reviewer: Neurips 2024, ICLR 2025, ICML 2025, Neurips 2025, Reliable ML workshop @ Neurips 2025, ICLR 2026.Emergency reviewer: COLM 2025, FORC 2025.
TA: Discrete Methods in CS (Spring 2025)
Student Activities
The first year of my PhD, I served as a CS department senator within the USC Viterbi Graduate Student Association (VGSA), where I was able to bring goats to campus (amongst other things).
I served as President / VP for my undergrad’s ACM chapter, where I am most proud to have established a perpetual $30k endowed scholarship with club funds. Importantly, the scholarship may also grant eligible students an in-state tuition waiver! Students can apply here.
I also served as the undergrad representative on the UTD CS department head search committee during the 2020-2021 academic year.
If you are a UTD student interested in pursuing a PhD, you may find the grad school section I wrote of this guide helpful. Also check out this research internship / REU masterlist which I put together.
Fun
In my free time I enjoy climbing/bouldering, surfing, hiking, and skiing (the latter two less frequently!). I have (at some point in the past) also attempted the following: learning art, running, calisthenics progressions, playing classical piano, swimming, and binging npr tiny desk concerts.
Some other stuff from undergrad:
- Clustering in hyperbolic space (with Mike Skinner)
- Short notes on statistical learning (with Dipnil Chakraborty)
- Point packing in hypercubes
- Coarse classification of russian troll tweets
Rejections
Here is a list of some of my rejections (e.g. jobs, fellowships, etc.) to help normalize the fact that most things don't work out :DGive, even if you only have a little.
-Siddhartha Gautama