Questions I keep coming back to

Research

How can learning systems recognize what they know, discover what they don’t, and remain useful when the world changes?

01 Generalization & open worlds

Recognizing known concepts while discovering new categories in unfamiliar domains. I work on representations and adaptation strategies that connect domain generalization with generalized category discovery.

DG-GCD ↗ HiDISC ↗
02 Geometry & structure

Hierarchies are everywhere—from geographic scenes to visual categories. I explore hyperbolic representations that make room for these structures, including open-world remote sensing.

HElIOS ↗
03 Language & multimodal learning

How we represent language affects what models retain and how they behave. My work spans morphology-aware Indic tokenization, multilingual interventions, and medical question answering across text and images.

SuTRA ↗ Beyond Accuracy ↗
04 AI in financial systems

At Motilal Oswal, I study structured financial generation, time-series representations, and the implications of agents sharing foundation models. Only public-safe research details are shared here.

CandleDiff ↗ Many Agents, Few Minds ↗

A little geometric intuition

Interactive notebook

In a Euclidean plane, circles grow linearly with radius. In a hyperbolic plane, their circumference grows exponentially at large distances—one reason to explore this geometry for branching hierarchies.

Euclidean plane: a circle at distance 2 from the origin
Euclidean planeCircumference: 12.57
Poincaré disk: a circle at hyperbolic distance 2 from the origin
Poincaré diskCircumference: 22.79

An illustration, not experimental results. Curvature −1; circumferences are 2πr and 2πsinh(r). The disk maps distance r to radius tanh(r/2); its boundary is infinitely far away. Screen sizes are not metric distances.

Ongoing research

Current questions and collaborations. Submission venues, review details, and private manuscript links are intentionally omitted.

Scalaris: A Scalable Enterprise Memory Architecture for Conversational AI Assistants

Devarsh Gandhi, Sarthak Pawar, Hemprasad Yashwant Badgujar, Vaibhav Rathore, Siddharth Sureka, Neha Bhargava

Develops a unified long- and short-term memory architecture for accurate, token-efficient retrieval in enterprise conversational systems.

Zero-Shot or Fine-Tuned? Benchmarking Time-Series Foundation Models Across Indian and US Equity Markets

Ritik, Vaibhav Rathore, Nachiketa Ratnakar Patil, Siddhant Gole, Siddharth Sureka, Neha Bhargava

Benchmarks zero-shot, last-block, and full fine-tuning strategies for time-series foundation models across Indian and US equity data.

NOMAD: Federated Generalized Category Discovery Across Unseen Domains

Vaibhav Rathore, Divyam Gupta, Mainak Singha, Neha Bhargava, Subhasis Chaudhuri, Elisa Ricci, Biplab Banerjee

Studies known- and novel-category discovery in unseen domains when source data remains decentralized and non-IID.

FoCUS: Bridging Fine-Grained Recognition and Open-World Discovery across Domains

Vaibhav Rathore, Divyam Gupta, Moloud Abdar, Subhasis Chaudhuri, Biplab Banerjee

Connects fine-grained recognition with open-world discovery under cross-domain distribution shifts.

Hyperbolic Geometry for Remote Sensing: From Hierarchical Scene Embeddings to Open-World Earth Observation

Vaibhav Rathore and collaborators

Develops a unified perspective on hyperbolic learning for hierarchical and open-world Earth observation.

Earlier explorations

AI Support Assistant

A retrieval-augmented support system built around product manuals and a QLoRA-tuned language model.

Kinector

Text-conditioned 2D gesture generation using MediaPipe keypoints and a GRU pose model.

Medical Image Deblurring

A scale-recurrent network with spatial-asymmetric attention for multimodal medical images.

Learning with Noisy Labels

Vision Transformer experiments on CIFAR-100 with substantial label noise.