Autoscience Institute builds AI systems that autonomously conduct AI research, including Carl, the first AI system to produce a peer-reviewed paper, accepted at ICLR 2025 Workshops.
Autoscience Institute builds AI systems that autonomously conduct AI research - in effect, automating the role of the AI research scientist itself. Based in San Mateo, California, the company pursues a single, unusually ambitious proposition: a research process in which machine learning models are created, evaluated and deployed with minimal human intervention, removing human bottlenecks from scientific discovery.
The company's flagship system, Carl, holds a distinction with few precedents: it is the first AI system to produce an academically peer-reviewed research paper, accepted at ICLR 2025 Workshops. Alongside Carl sits an autonomous lab, an environment in which AI systems develop, assess and ship new machine learning models end to end. The technical work spans deep learning, reinforcement learning, genetic algorithms, distributed model training and machine learning model development.
Autoscience Institute has raised $14 million in seed funding, led by General Catalyst with participation from the Perplexity Fund and Toyota Ventures. The founding team brings backgrounds from Alphabet's X, MIT and Harvard. The company operates at the intersection of AI research, scientific discovery and machine learning - work that is measured by peer-reviewed results rather than benchmarks alone.
