ईक्षा
pronounced "eek-sha" · noun (Sanskrit)
the act of seeing; a directed gaze — adopted here as a research orientation toward examining the internal structure of neural computation.
Iksha is an emerging research lab working on mechanistic interpretability in artificial intelligence.
We study how neural networks compute internally, focusing on the structure of learned representations rather than surface behavior. Our work aims to develop methods for inspecting, reasoning about, and answering open-ended questions concerning how these systems organize and transform information.