Aditya Gautam is an AI engineer and leader working at the intersection of large language models, multi-agent systems, and their application to recommendation and content understanding. At Meta, he has led several initiatives applying LLMs at scale across recommendation and ranking, delivering multiple user engagement wins. He architected a novel multimodal LLM for Reels understanding and owned its full training stack — continued pre-training on in-domain video and multimodal signals, mid-training for task and policy adaptation, and post-training through distillation, DPO, and adaptive retrieval — then extended it to the hardest adversarial problems in short-form video, including modality manipulation and content originality attribution. Alongside this, he has been a central contributor to Meta's multi-agent system design and to optimizing large-scale training and serving workflows for cost and throughput. Prior to Meta, Aditya spent more than three years at Google working in computer vision.

His research covers multimodal understanding with LLMs, multi-agent systems, and LLM evaluation, with publications at top-tier venues including ICWSM and WSDM. His leadership in the field has been featured in interviews with Nasdaq and technical publications such as Marktechpost and Uniquely Wired. A highly active voice in the generative AI community, he frequently presents at major industry events like the Databricks Data + AI Summit and serves as a peer reviewer and ethics reviewer for top-tier academic conferences including NeurIPS, ICML, and AAAI.