Hi, I'm Srinitish Srinivasan
Junior Research Fellow(JRF) at Indian Institute of Technology, Delhi researching Geometric Deep Learning.
Junior Research Fellow(JRF) at Indian Institute of Technology, Delhi researching Geometric Deep Learning.
I am a final-year undergraduate student in Computer Science at Vellore Institute of Technology, specializing in representation learning, particularly on Geometric Deep Learning, Out-Of-distribution(OOD) generalization and Riemannian Machine Learning. I also have a keen interest in State Space Models and Agent based Systems. Currently I am a Junior Research Fellow(JRF) at LCS2, IIT Delhi under Dr.Tanmoy Chakraborty.
My final-year undergrad research was focused on studying the discriminative bottleneck of Hyperbolic GNNs and developing an approximate solution to tackle the same. Previously, I was a visiting research student at the University of Lincoln, where I worked on mitigating spurious correlations in object detection under the guidance of Dr. Karthik Seemakurthy.
Received a travel grant of USD 1500 from GRADES NDA, ACM SIGMOD/PODS to present my paper LGIN: Easing the Injectivity Bottleneck in Hyperbolic GNNs
June, 2025
My paper LGIN has been accepted at ACM SIGMOD 2025 GRADES NDA Workshop
May, 2025
Joined LCS2, IIT Delhi as a Research Fellow under Dr.Tanmoy Chakraborty.
May, 2025
My paper on Joint Predictive and Bayesian Inference based Graph Self Supervised Learning is out on Arxiv.
February, 2025
Srinitish Srinivasan, Omkumar CU. ACM SIGMOD 2025 GRADES NDA Workshop
Hyperbolic GNNs may not be as powerful as Euclidean GNNs due to non-injective aggregations. LGIN provides a theoretical framework for a Hyperbolic GNN that approximates a powerful GNN despite constraints on Hyperbolic spaces.
J. S. Walia, A. Sinha, S. Srinivasan, S. Unnikrishnan. Preprint
Designed a CausalGAN and RL-based framework to generate synthetic bond yield data. Integrated a fine-tuned LLM to provide trading signals, risk assessments, and volatility projections.
Srinitish Srinivasan, Omkumar CU. Scientific Reports
A white-box gradient-based adversarial attack which targets the contrastive latent space, demonstrating its effectiveness by reducing model performance by an average of 20-25% on tested graph classification tasks.
I'm always interested in discussing research collaborations, speaking opportunities, freelancing, or just chatting about AI/ML. Feel free to reach out!
smudge0110@icloud.com
Discord
smudge0110
Office
LCS2, Yardi School of AI, IIT Delhi
Book a 1:1 video call to discuss collaborations or opportunities. Please mention the purpose of the call in the meeting details.