NeoSigma

NeoSigma is a product-driven research lab based in San Francisco, CA, that builds what it describes as the intelligence layer linking customers, products, and AI systems. Its technical work concentrates on machine learning infrastructure for AI agents, with a specific focus on enabling agents to learn from experience in production and on evaluation and reliability of production AI systems.

The company's main product is an agent learning infrastructure layer. It captures agent failures, converts them into structured evaluation signals, and feeds those signals back into continuous improvements in agent reliability. NeoSigma states that this process compounds into proprietary intelligence across users, teams, and organisations. Relevant domains include feedback loop systems, agent learning from experience, and frontier-scale model training.

NeoSigma is a small team of researchers and engineers; members have trained frontier-scale models and built widely used AI products and agents. Team backgrounds include MIT, Parallel Web, Essential AI, Apple, and Amazon. The company is backed by investors and leaders from Google DeepMind, OpenAI, World Labs, Mercor, and Decagon.

The firm reports being SOC 2 Type 2 and HIPAA compliant. It operates in the artificial intelligence, AI infrastructure, and developer and ML tooling sectors.