Sureform

We collect high-quality multimodal human data, starting with speech and video across diverse expressions, emotions, and interactions, to help companies build AI models that engage more naturally with the world.

High-Level Overview

Sureform is a startup specializing in collecting high-quality multimodal human data, primarily video and speech, to train AI models that interact more naturally with the physical world. Their data captures diverse human expressions, emotions, and everyday activities, enabling advancements in multimodal and physical AI such as humanoid robots and video generators[1][2][3]. Founded in 2025 and based in San Francisco, Sureform serves AI companies and researchers who need rich, realistic datasets to improve AI understanding and interaction capabilities. The company addresses the critical problem of limited high-quality, diverse human data for training AI, which is essential for building more natural and effective AI systems. Early traction includes participation in Y Combinator’s Spring 2025 batch and active data collection roles involving everyday task recordings[1][3].

Origin Story

Sureform was founded in 2025 by Ananth Kashyap, who previously worked at Google and studied at the University of Pennsylvania. The idea emerged from the need for richer, more diverse human data to train AI models that can better understand and engage with the world in a multimodal way (combining video, audio, and other sensory inputs). The company quickly gained momentum by focusing on capturing first-person point-of-view video of daily tasks, a niche that supports training robotics and AI models for real-world applications[1][3].

Core Differentiators

• High-quality multimodal data: Sureform collects synchronized video and audio data capturing natural human interactions and tasks from multiple angles, including POV footage.
• Focus on physical and multimodal AI: Unlike many datasets that focus on text or isolated modalities, Sureform emphasizes rich, real-world data for embodied AI systems.
• Diverse task coverage: Their data includes everyday household and workshop activities such as cooking, cleaning, painting, and repairs.
• Flexible data capture setups: Contributors use various cameras (GoPro, Meta Ray Ban Glasses, smartphones) and multi-camera rigs to provide comprehensive perspectives.
• Early integration with AI training: The data directly supports training advanced AI models for robotics and video generation, accelerating development cycles[1][3].

Role in the Broader Tech Landscape

Sureform rides the growing trend of multimodal AI and embodied intelligence, where AI systems integrate vision, audio, and physical interaction to operate in the real world. The timing is critical as AI moves beyond text and static images toward understanding complex human environments and behaviors. Market forces such as increased demand for humanoid robots, AI assistants, and video synthesis technologies favor companies like Sureform that provide foundational data. By enabling more natural AI-human interaction, Sureform influences the broader ecosystem by filling a key data gap that many AI developers face, potentially accelerating breakthroughs in robotics, AR/VR, and interactive AI applications[1][2][3].

Quick Take & Future Outlook

Looking ahead, Sureform is poised to expand its dataset scope and scale, potentially incorporating more diverse environments and interaction types. As multimodal AI and physical AI applications grow, demand for high-quality human data will increase, positioning Sureform as a critical infrastructure provider. Trends such as robotics automation, AI-driven video content creation, and augmented reality will shape their journey. Their influence may evolve from a data provider to a strategic partner for AI companies building next-generation interactive systems, helping to set standards for multimodal data quality and diversity[1][3]. This aligns with their mission to enable AI models that engage more naturally and effectively with the world.