AI-native code vulnerability detection platform for developers
Depthfirst builds an AI-native security scanner that identifies code vulnerabilities while filtering false positives—a critical friction point in developer security workflows. The tech stack reveals a modern ML-first architecture: PyTorch for model training, Temporal for complex agent orchestration, and TypeScript/Python across frontend and backend. Active projects center on building agentic AI pipelines for vulnerability discovery and scaling proof-of-concept exploitation, while the hiring surge (11 roles in 30 days, mostly senior) signals aggressive scaling of both security research and sales—a tell-tale sign of a pre-launch or Series A company.
Depthfirst is a San Francisco-based security startup (11–50 employees) built for engineering teams at mid-market and enterprise companies. The platform uses AI agents to analyze source code, business logic, and infrastructure patterns to surface exploitable vulnerabilities and deliver fixes directly in developer workflows (GitHub, GitLab, Jenkins). The company is solving two acute problems: false-positive noise that makes traditional SAST tools unusable, and the difficulty of discovering novel or AI-enabled zero-day vulnerabilities at scale. Infrastructure and product engineering are undersized relative to sales and research hiring, suggesting the platform core is maturing and go-to-market is the next frontier.
TypeScript, Python, React, Next.js, PostgreSQL, Redis, and AWS on infrastructure. PyTorch powers model training, Temporal handles agent orchestration, and Neon, Redux, Zustand, and Tailwind CSS support the frontend.
An AI-native platform that scans code and infrastructure to find real vulnerabilities, reduce false positives, and deliver actionable fixes into developer workflows. Uses agentic AI pipelines and proof-of-concept exploitation to uncover novel and zero-day vulnerabilities.
Other companies in the same industry, closest in size