Research

Self-Distillation without Imitation - Bootstrapping Dense Reward Models from Privileged Information

Work in Progress (2026)

Contrastive ground-level image and remote sensing pre-training improves representation learning for natural world imagery

Andy V. Huynh, Lauren E. Gillespie, Jael Lopez-Saucedo, Claire Tang, Rohan Sikand, Moisés Expósito-Alonso

Beyond the Imitation Game - Quantifying and extrapolating the capabilities of language models

Aarohi Srivastava, Abhinav Rastogi, Abhishek Rao, et al. (incl. Rohan Sikand)

Parallax - Agentic Satellite Query Engine

An AI agent that answers natural-language questions about Earth using live satellite imagery.

tinyrl

A minimal reinforcement learning toolkit — simple environments, policies, and training utilities.