Obsidian Labs closes $85M to build private LLMs for product design teams

AI · 5 min read

Obsidian Labs announced an $85 million Series B led by Accel to accelerate work on private large language models for product and design teams. The company positions its models as enterprise-safe assistants that can generate component code, produce design rationale, and summarize user research while keeping data on customer-owned infrastructure.

Obsidian's platform offers a training pipeline that ingests design systems, Figma files, product documentation, and anonymized research transcripts to create a team-specific model. The company highlighted features such as an "explainable rationale" mode that traces recommendations back to source docs and component tokens.

Funding will be used to expand compliance certifications (SOC2, ISO), improve on-device inference options for sensitive projects, and build deeper integrations with Atlassian, Notion, and major design tools. Customer pilots include regulated industries where data residency and auditability are required.