Profile Picture

Sree Harsha Tanneru

Graduate Student Researcher @ Harvard University

About Me

Hey there! Welcome to my personal website. I am a 2nd year Masters (Thesis) student in Computational Science at Harvard University advised by Prof. Himabindu Lakkaraju. I work broadly in the realm of trustworthy machine learning, and particularly, explainability, uncertainty quantification, interpretability, and fairness.

Prior to graduate school, I was a Lead Product Engineer in Machine Learning at Sprinklr AI for three years developing distributed machine learning systems for vision, language and speech processing. A few products I worked on - Visual Insights, Conversational AI, Voice AI. In what seems like a lifetime ago, I graduated with a Bachelors degree from Indian Institute of Technology Madras where I was advised by Prof. Balaji Srinivasan. I had a fun time working on problems in computational fluid flow simulations and numerical methods, which gradually turned into an interest in machine learning.

Find Me

email    twitter    github    google scholar    linkedin    goodreads   

Research Interests

  1. Faithfulness in foundation models
  2. Explainability
  3. Interpretability

Publications

  1. Chirag Agarwal, Sree Harsha Tanneru, Himabindu Lakkaraju. Faithfulness vs. Plausibility: On the (Un)Reliability of Explanations from Large Language Models, Preprint. Under Review.

  2. Himabindu Lakkaraju, Sree Harsha Tanneru*, Chirag Agarwal*. Uncertainty In Natural Language Explanations of Large Language Models, Accepted at AISTATS 2024, and a Spotlight Paper at NeurIPS R0-FoMo 2023 Workshop

  3. Alexander Lin*, Sree Harsha Tanneru*, Lucas Monteiro Paes*, Suraj Srinivas, and Himabindu Lakkaraju. Word-Level Explanations for Analyzing Bias in Text-to-Image Models, Presented at ICML 2023 Workshop on Challenges of Deploying Generative AI

  4. Sree Harsha Tanneru*. Edge Importance Scores for Editing Graph Topology to Preserve Fairness, Presented at ICML 2023 Workshop on Topology, Algebra, and Geometry in Machine Learning

  5. Dr. Balaji Srinivasan, T Sree Harsha*. Reconstruction of Flows using Convolutional Neural Networks, ISME Journal of Thermofluids Volume 5(1): 01-15. doi: 112019

Reviewing

Teaching Experiences

I had the fortune of being a teaching fellow for following classes during my masters.

Here's what my students have to say about me. Not so subtle *humble brag* follows.
word cloud
  • "Sree was extremely helpful during office hours in explaining more complex topics discussed in class."
  • "Absolutely stellar TF, very grateful for his help during the course."
  • "Super helpful when I was confused about something!"
  • "He was an invaluable TF for so many of us and helped us whenever we had questions or concerns."
  • "He was extremely nice with everybody and made sure every single person fully understood the concepts and had their questions answered in their entirety."
  • "Couldn't have gone through 187 without him!"
  • "I wasn't able to go to Sree's office hours until about halfway through the semester, and I really regret that I couldn't go sooner."
  • "His office hours were very helpful in understanding the material but he also answered our (usually dumb) questions in such a respectful way that encouraged us to ask more questions so that we could fully comprehend the material."
  • "One of the best tf's i've had in a class. Would not have made it through this class without Sree."

Books

I (make an attempt to) read. Find me on goodreads.

Currently reading

Next up