Vellum

Vellum is an AI product development platform providing prompt management, model comparison, workflow orchestration, and production monitoring tools for engineering and product teams building LLM-powered applications — enabling teams to iterate on AI features with rigorous evaluation frameworks rather than ad-hoc prompt tweaking. Founded in 2023 by Andrew Kirima and Noa Flaherty in San Francisco, Vellum has raised approximately $12 million and targets AI-forward product teams at growth companies who need structured workflows for LLM feature development, testing, and deployment.\n\nVellum's platform covers the LLM application development lifecycle: Prompt Workshop for managing and versioning prompt templates with variable substitution, Evaluations for testing prompts against datasets to measure output quality before deployment, Document Index for building RAG (retrieval-augmented generation) pipelines with semantic search over enterprise documents, and Workflows for orchestrating multi-step AI pipelines with branching logic and human-in-the-loop review steps. The monitoring dashboard tracks production LLM performance, latency, and cost across models.\n\nIn 2025, Vellum competes in the rapidly growing LLM development tools market against LangSmith (LangChain's commercial platform), Weights & Biases Prompts, Helicone, Braintrust, and Humanloop for AI application observability and evaluation. The market has grown explosively as companies productionize LLM features and need rigorous quality control processes. Vellum's differentiation is its end-to-end workflow — from prompt development through evaluation to production monitoring — in a single platform rather than requiring separate tools for each stage. The 2025 strategy focuses on expanding workflow complexity support (longer multi-agent pipelines), growing enterprise adoption with SSO and access controls, and adding AI-powered evaluation that automatically judges output quality.