
Ex-Meta AI Execs Raise $15M for AI Assistant Startup Yutori

Two ex-Meta artificial intelligence executives have raised 15 million to start Yutori, a startup that aims to build AI-powered personal assistants, the company said on Thursday.
The round was led by Rob Toews at Radical Ventures, with participation from investors such as Felicis, AI pioneer Fei-Fei Li, and Google DeepMind chief scientist Jeff Dean.
San Francisco based Yutori is part of an expanding trend of AI companies focused on autonomous agents AI-powered systems that can carry out tasks independently. OpenAI CFO Sarah Friar and other industry leaders have highlighted that such systems will be at the forefront this year, given that AI models have become sophisticated enough to perform intricate, step-by-step tasks online with little need for human intervention.
"Everything is happening in the space around chatbots, but they aren't really alleviating tasks from your plate," Yutori co-founder Devi Parikh told Reuters in an interview. She said Yutori is rethinking users' interactions with AI-powered agents to make more efficient work out of tasks like ordering food, or handling complicated travel plans.
The startup is investing heavily in post-training methods to enhance AI models' capability to surf the web and optimize their performance for specific tasks. Post-training has emerged as a dominant factor in improving AI reasoning models, including OpenAI's o1 and o3 models.
Yutori’s leadership team includes Parikh, who previously led multimodal AI research at Meta , and Dhruv Batra, who headed Meta's embodied AI research, working on AI models for robotic navigation in 3D environments. Other team members include experts responsible for post-training Meta's flagship open-source models, Llama 3 and Llama 4.
The round was led by Rob Toews at Radical Ventures, with participation from investors such as Felicis, AI pioneer Fei-Fei Li, and Google DeepMind chief scientist Jeff Dean.
San Francisco based Yutori is part of an expanding trend of AI companies focused on autonomous agents AI-powered systems that can carry out tasks independently. OpenAI CFO Sarah Friar and other industry leaders have highlighted that such systems will be at the forefront this year, given that AI models have become sophisticated enough to perform intricate, step-by-step tasks online with little need for human intervention.
"Everything is happening in the space around chatbots, but they aren't really alleviating tasks from your plate," Yutori co-founder Devi Parikh told Reuters in an interview. She said Yutori is rethinking users' interactions with AI-powered agents to make more efficient work out of tasks like ordering food, or handling complicated travel plans.
The startup is investing heavily in post-training methods to enhance AI models' capability to surf the web and optimize their performance for specific tasks. Post-training has emerged as a dominant factor in improving AI reasoning models, including OpenAI's o1 and o3 models.
Yutori’s leadership team includes Parikh, who previously led multimodal AI research at Meta , and Dhruv Batra, who headed Meta's embodied AI research, working on AI models for robotic navigation in 3D environments. Other team members include experts responsible for post-training Meta's flagship open-source models, Llama 3 and Llama 4.