About Me
GitHub / LinkedIn / Google Scholar / Email
Welcome to my homepage.
Hello! I’m Poornash Anandan Sangeetha (often going by A S Poornash), an M.S. student in Machine Learning at Carnegie Mellon University (CMU), USA (2025-2027). I will also be working as an ML Intern at Tiktok’s Recommendation algorithm team during the summer of 2026 in the Bay Area.
I also graduated as the Institute Topper (President of India’s Gold Medalist) in Computer Science and Engineering from the Indian Institute of Technology Patna (IIT Patna), with a GPA of 9.71/10.
I served as a Machine Learning Intern at Amazon Applied Science (Jan-Jul 2025), where I developed automated workflows for vision-language product matching, achieving over 90% reduction in manual review time and significant gains in classification metrics. Before this, I worked as a researcher in the ODDS Research Group at the University of Southern California (USC), advised by Prof. Meisam Razaviyayn, focusing on private, fair, and federated machine learning, as part of the competitive IUSSTF-Viterbi India Scholar program.
At IIT Patna, I completed my undergraduate thesis under Prof. Sriparna Saha and worked on research problems in NLP and multimodal learning (including projects accepted at conferences and journals like ACL, ACM Multimedia, AAAI, CIKM and, IEEE TCSS). I also collaborated with Prof. Kitsuchart Pasupa (KMITL, Thailand) on hate speech detection for low-resource languages like Thai. Further details and papers are available on my Publications page.
My research interests span natural language processing, multimodal learning, and machine learning, with a focus on building trustworthy and fair systems. I have worked on advancing methods for toxicity detection, addressing challenges in multilingual and low-resource language settings, and exploring effective multimodal alignment methods. My work includes contributions to privacy-preserving and fair learning in decentralized environments, as well as improving readability and relevance in summarization systems. Additionally, I am interested in the intersection of vision and language, particularly in tasks requiring robust cross-modal alignment and reasoning.
Some relatively non-academic things about me -
- I am an avid puzzle solver and pursue speedcubing in a competitive manner. (My WCA profile link)
- Quite a philosopher myself. Speak 4 languages.
Please read my CV or visit Scholar profile to learn more about my work and interests.
News
January 2025: Started my internship as an Applied Scientist Intern at the Amazon Global Store Tech team in Noida, India. Curious to explore the industrial side of ML!
December 2024: An initial version of my paper titled “A Stochastic Optimization Framework for Private and Fair Learning From Decentralized Data” was accepted at the PPAI-Workshop @ AAAI-2025! This work was a result of my research fellowship at USC.
October 2024: My team, DBkaScam, placed 6th among 20,000+ teams in the prestigious Amazon ML Challenge 2024! Check out our winning solution.
July 2024: Paper titled “Seeing Beyond Words: Multimodal Aspect-Level Complaint Detection in Ecommerce Videos” accepted at ACM MM 2024 as an Oral Presentation (<3% acceptance rate).
June 2024: Paper titled “ToxVI: A Multimodal LLM-based Framework for Generating Intervention in Toxic Code-Mixed Videos” accepted at CIKM 2024. This extends our earlier ACL work to further mitigate the spread of toxic posts on social media.
May 2024: Started my internship at the ODDS Research group at USC as one of the 15 IUSSTF-Viterbi Research Scholars selected from across India. Looking forward to a great summer in Los Angeles!
May 2024: Paper titled “ToxVidLM: A Multimodal Framework for Toxicity Detection in Code-Mixed Videos” accepted at the Findings of ACL 2024. See you in Thailand!
November 2024: Paper titled “HateThaiSent: Sentiment-Aided Hate Speech Detection in Thai Language” accepted at IEEE Transactions on Computational Social Systems (IEEE TCSS). This work was done as part of my research internship at KMITL, Thailand.
May 2023: Paper titled “APTSumm at BioLaySumm Task 1: Biomedical Breakdown, Improving Readability by Relevancy Based Selection” accepted to the BioNLP Workshop @ ACL 2023. This was my first-ever research project.