Vellum

Vellum is a platform to manage the full lifecycle of LLM-powered apps, letting users select the best prompts and models, compose them with RAG and APIs, and deploy and iterate on them in production. The platform includes a suite of tools for prompt engineering, semantic search, prompt chaining, evaluation, and monitoring, making it easier to get LLM features from prototype to production.

Vellum supports a broad variety of use cases, including workflow automation, document analysis, copilots, fine-tuning, Q&A over documents, intent classification, summarization, vector search, email generation, chatbots, blog generation, topic summarization, sentiment analysis, model distillation, and data extraction. The platform is designed to let both technical and non-technical teams experiment with new prompts and models without affecting production.

Some of the key features of Vellum include:

• Prompt Engineering: Experiment with new prompts and models without affecting production.
• Compose Complex Multi-Step Chains: Build complex chains of prompts and business logic with powerful versioning, debugging, and monitoring tools.
• Out-of-Box RAG: Upload text and search across text with a simple endpoint, no infrastructure required.
• Evaluate Prompts at Scale: Quantitatively evaluate with popular metrics or custom definitions.
• Deploy Changes with Confidence: Includes Github-style release management and Datadog-style monitoring for safe deployments.

Vellum is designed for enterprise scale, with best-in-class security, privacy, and scalability. It is SOC2 Type II Compliant, HIPAA Compliant, and offers Virtual Private Cloud deployments with support from AI experts and customizable data retention and access.

Real-world examples of Vellum in use include Drata, where cross-functional teams work together to rapidly iterate on new AI use cases, and Redfin, which used Vellum's Evaluation framework to simulate hundreds of real-world test cases for its virtual assistant before launch. Lavender's engineering team also uses Vellum to continuously refine live LLM features, adapting to production edge cases and handling over 90k requests monthly.

Vellum is funded by top VCs including Y Combinator, Rebel Fund, Eastlink Capital, and the founders of HubSpot, Reddit, Dropbox, Cruise, and Instacart, a strong vote of confidence in the potential of Vellum as a powerful and reliable way to build LLM applications. Interested parties can request a demo to see how Vellum can help them bring their AI ideas to production quickly and efficiently.

Published on June 14, 2024