I can point out sections. Click me.

Nitish
Srinitish Srinivasan

Machine learning research across geometry and representation learning

Master's student at IIT Delhi, advised by Dr. Tanmoy Chakraborty.

Get In Touch View Research View Projects

About Me

I am a Master's Student at Indian Institute of Technology, Delhi supervised by Dr.Tanmoy Chakraborty specializing in Information Retrieval and Geometry. Over the past 1 year, I have been working on geometric representations of hierarchies such as Taxonomies. I also have a keen interest in agentic memory.

My final-year undergrad research was focused on studying the discriminative bottleneck of Hyperbolic GNNs and developing an approximate solution to tackle the same. I'm always excited to collaborate at the intersection of geometry, graphs, information and agentic retrieval. I also take freelance projects in cloud native and applied AI systems. If any of this resonates with you, feel free to reach out. I'm always happy to chat about new ideas and projects.

Apart from research, I enjoy cycling, running, K-dramas, Korean music (massive Uaena btw :p), and travelling.

Geometric Deep Learning
Graphs
Taxonomy Induction
Download CV

Recent News

Paper Accepted

Our paper on Coupled Orbital Polar Embeddings for learning hierarchies has been accepted to ICML 2026. See y'all in Seoul this summer.

May, 2026

Paper Accepted

Our paper on Gaussian Box Embeddings for Taxonomy expansion has been accepted at ACM WebConf 2026 Main Track. See y'all in Dubai this April hopefully :)

January, 2026

Reviewer

Serving as a PC member for WWW Tracks.

January, 2026

University

Began my Master's program at IIT Delhi under the supervision of Dr. Tanmoy Chakraborty.

January, 2026 (Spring Intake)

Conference

Attended ACM Sigmod in Berlin, Germany and presented my poster at GRADES NDA Workshop.

June, 2025

Grant Awarded

Received a travel grant of USD 1500 from GRADES NDA to present my poster at ACM SIGMOD.

June, 2025

Paper Accepted

My undergraduate thesis paper has been accepted at ACM SIGMOD 2025 GRADES NDA Workshop.

May, 2025

Research

Recent Publications

Selected work on hierarchy, graph geometry, self-supervision, and causal modeling.

Polar embedding visualization for Polaris
ICML 2026 Hierarchy Learning

Polaris: Coupled Orbital Polar Embeddings for Hierarchical Concept Learning

Srinitish Srinivasan*, Sahil Mishra*, Srikanta Bedathur, Tanmoy Chakraborty.

A polar hyperspherical embedding framework for structured knowledge representations across taxonomies and ontologies. Polaris disentangles semantic meaning from hierarchy using angular and radial geometry.

Paper Code
TaxoBell model diagram
WebConf 2026 Taxonomy Expansion

TaxoBell: Gaussian Box Embeddings for Self-Supervised Taxonomy Expansion

Sahil Mishra, Srinitish Srinivasan, Srikanta Bedathur, Tanmoy Chakraborty.

A Gaussian box embedding framework for asymmetric taxonomy relationships, where means encode semantic location and covariances encode uncertainty across single and multi-parent taxonomies.

Paper Code
LGIN poster preview
SIGMOD 2025 Workshop Poster

Can we ease the Injectivity Bottleneck on Lorentzian Manifolds for GNNs?

Srinitish Srinivasan, Omkumar CU.

LGIN studies non-injective aggregations in hyperbolic GNNs and provides a theoretical framework for approximating powerful GNN behavior under hyperbolic space constraints.

Paper Poster Code
Causal GAN financial modeling architecture
Preprint Financial Modeling

Predicting Liquidity-Aware Bond Yields using Causal GANs

Srinitish Srinivasan*, Karan Walia*, Aarush Sinha*, Srihari Unnikrishnan*.

A CausalGAN and reinforcement-learning framework for synthetic bond yield data, trading signals, risk assessment, and volatility projection.

Paper Code
Graph adversarial attack visualization
Scientific Reports Adversarial GNNs

An edge sensitivity based gradient attack on Graph Isomorphic Networks

Srinitish Srinivasan, Omkumar CU.

A white-box gradient adversarial attack based on contrastive latent-space embeddings for graph neural networks, validated across four benchmark datasets.

Paper Code
JPEB pipeline diagram
Preprint Self-Supervised GNNs

Predict, Cluster, Refine: A JEPA Framework for GNNs

Srinitish Srinivasan, Omkumar CU.

A graph self-supervised learning technique that combines joint predictive embedding architecture with Gaussian kernels for robust representation learning.

Paper Code
View All on Google Scholar
Engineering

Development Projects

Systems work for agents, search, ML deployment, and cloud workflows.

Multi-Agent Systems Protein Misfolding

Multi-Agent System for Protein Misfolding Analysis

Agent workflow for protein misfolding analysis with hypothesis refinement and biological interaction checks.

SageMaker-backed experimentation workflows Inter-agent coordination protocols
Multi-Agent Systems AWS AWS SageMaker Lambda API Gateway Cognito/Auth
Agentic Search Fashion Retrieval

Agentic Fashion Search API

Semantic and intent-driven fashion retrieval across product catalogs, exposed as an API.

Fargate services SQS task queue DynamoDB storage
AWS AWS Fargate API Gateway SQS DynamoDB Gemini API JWT
Mobile AI Conversational Interface

Plants-to-Human Conversational Interface

Mobile interface for plant-state conversations with generative AI and cloud-backed login.

Flutter cross-platform frontend Gemini-powered contextual dialogue
Flutter GCP Firebase Auth Gemini API Mobile

Get In Touch

Contact Information

Reach out for research, talks, systems work, or paper discussions.

Email

smudge0110@icloud.com

Discord

smudge0110

Office

LCS2, Yardi School of AI, IIT Delhi

Connect Directly on Discord

Discord QR Code

For quick paper or project discussions, Discord works too.

Username: smudge0110