Priya Goyal
Principal Research Engineer, LinkedIn CoreAI (leading Agentic Search)
ex-Google Deepmind, Founding Member DatologyAI, FAIR (Meta MSL)
Google Scholar (>65K citations)
LinkedIn
Twitter
Email
Github
I am currently a Principal Research Engineer at LinkedIn where I founded the Agentic Search expanding it into company priority and 50+ people effort. I architected a novel agentic evaluation framework adapted across the company. The eval framework enabled verifiable rewards for a very subjective problem. I also innovated the high quality grounded synthetic data generation to enable 0->1 product development. Also appointed lead by CTO and Chief AI Officer to standardize AI eval for the company.
Before that, I was a Founding Member of DatologyAI where I led company critical customer facing projects on data quality leading to more efficient and higher quality language models. Senior most researcher on the team informing clear research plans and priorities for the company including infrastructure investments. Before that, I was a Staff AI Researcher at Google Deepmind where I worked on building fundamental Multimodal technology aimed at enabling new capabilities with LLMs.
Prior to that, I spent almost 7 wonderful years at Facebook AI Research (FAIR) in New York, USA where I worked on computer vision and machine learning. At FAIR, I led research on self-supervised representation learning using self-supervision from uncurated datasets, training large-scale computer vision models (notable projects: ImageNet in 1 Hour and SEER 10 billion parameters self-supervised model) and building socially responsible AI models. I led the development of self-supervised learning library VISSL and am a recipient of the Best Student Paper Award for Focal Loss at ICCV 2017. I also led and organized the first ever self-supervised learning challenge at ICCV'19.
My research interests include agentic evaluations, high quality synthetic data generation, data curation / filtering, multimodal learning, computer vision and developing socially responsible AI.
