Vellum

Vellum AI is a development platform designed for building and deploying large language model (LLM) powered applications. The platform includes tools for prompt engineering, semantic search, model versioning, performance monitoring, and workflow automation. Through the platform, users can experiment with and switch between different LLM providers including OpenAI, Anthropic, Cohere, Google, MosaicML and hosted models like Falcon-40 billion-instruct and various Llama2 versions. The platform features a prompt engineering interface that enables users to design, iterate and test various prompts effectively while managing cost and performance.

Vellum's platform facilitates systematic iteration, data-driven experimentation, and regression testing before deployment. It also provides observability and monitoring capabilities that allow users to track metrics like quality, latency and costs over time. The platform supports automated testing, version control, and continuous integration/continuous delivery (CI/CD) practices. A key differentiator is its provider-agnostic approach, allowing businesses to avoid dependence on a single LLM provider.

The platform enables integration of company-specific context through semantic search capabilities without requiring users to manage their own semantic search infrastructure. Its workflow automation features help streamline processes and increase efficiency. For data privacy and security, Vellum offers configurable data retention and access controls to meet diverse privacy requirements, including HIPAA compliance for healthcare organizations.

Key customers and partnerships

The platform serves a diverse range of customers including tech startups, healthcare organizations, educational institutions, and customer support services.