Vedant Gaur is the Founder of Latent (YC F26), an AI reliability infrastructure company that reads a model's internal activations to detect and intercept wrong answers before they reach end users.

Vedant studied at the University of Pennsylvania's elite Jerome Fisher M&T Program, earning dual degrees in Computer Science and Economics/Finance (Wharton), along with a concurrent M.S.E. in Computer Science, while conducting NLP and interpretability research at Stanford and Princeton.

Vedant has published first-author research at ACL 2023 and EMNLP 2025 on LLM reasoning and benchmarking, and conducted mechanistic interpretability research at the Stanford NLP Group achieving ~98% AUROC on single-classification accuracy.

Prior to founding Latent, Vedant shipped production ML at Amazon and worked on AI at Robinhood, giving him firsthand experience with the trust and reliability challenges teams face when deploying open-weight models at scale.

Latent is part of Y Combinator's F26 batch and targets high-stakes industries including finance, healthcare, and legal, where a single confidently wrong model answer can create significant brand and liability risk.