The funding round, announced Thursday, was led by Rob Toews at Radical Ventures, with additional participation from investors including Felicis, AI pioneer Fei-Fei Li, and Google DeepMind’s Jeff Dean.
Yutori’s goal is “to transform how users interact with the web” by creating AI agents capable of performing tasks autonomously, from scheduling to managing finances.
Parikh and Das left Meta a year ago to start the company, with Batra joining shortly after. Their focus is on overcoming the limitations of current chatbots, which struggle to handle complex tasks without human intervention.
“Productivity isn’t about cramming more into your day — it’s about reclaiming your attention for what truly matters,” said Devi Parikh, co-CEO and co-founder of Yutori. “Yutori’s mission is to build AI assistants that make space for the meaningful things in life.”
The startup’s approach centers on “post-training” models, which enhance AI systems’ ability to navigate the web. By refining these models after initial training, Yutori seeks to improve their accuracy in completing long and complex actions.
The co-founders bring decades of combined AI research experience. Devi Parikh led multimodal AI research at Meta, contributing to projects like Llama 3, Make-A-Scene, and Emu.
Dhruv Batra previously led Meta’s embodied AI research, focusing on systems that enable robots to interact with the physical world.
Abhishek Das, a former research scientist at Meta’s Fundamental AI Research (FAIR) team, worked on deep neural networks and their applications in climate change.
He was advised by Batra and collaborated closely with Parikh while pursuing his PhD in computer science at Georgia Tech. His thesis on building agents that can see, talk, and act earned recognition.
Yutori’s team is designing systems where multiple AI agents work in parallel, learning from each other to improve overall efficiency. The goal is to unlock “superhuman” performance, enabling AI systems to execute tasks with high resilience and accuracy.
Yutori’s approach challenges the conventional method of building AI systems, which often rely on large language models (LLMs). Instead, the startup focuses on a holistic agent-first design.
“The web is the ultimate digital environment, but it’s also noisy and dynamic, which means mistakes are inevitable,” said Batra. “The key agentic skill is resilience. Yutori’s agent-first approach will unlock superhuman performance.”
The company is also leveraging techniques like reinforcement learning, test-time search, and model-in-the-loop flywheels to enhance its agents’ performance. These innovations allow Yutori’s AI systems to autonomously complete tasks with fewer errors.
The $15 million seed funding will help Yutori expand its engineering and design teams and prepare for the launch of its products. The company plans to invite early adopters for a closed beta this spring, with a waitlist available on its website.
