Relari
LLM evaluation and testing framework
About
Founded in 2023 by Pasquale Antonante (MIT PhD in fault detection for AI systems) and Yi Zhang (Harvard MBA), Relari provides a testing and simulation stack for hardening LLM-based applications. Its open-source library continuous-eval offers modular metrics covering retrieval quality, text generation, code generation, classification, and agent behavior, enabling data-driven pipeline evaluation rather than ad hoc spot-checking. The platform also includes synthetic test data generation to stress-test GenAI applications before production. Relari targets AI engineering teams building RAG systems and autonomous agents in mission-critical settings, drawing on methodologies from autonomous vehicle simulation pipelines. The company is backed by Y Combinator (W24) and headquartered in Cambridge, MA.
Products
-
Nuvi
— AI agent builder that turns natural language specs into reliable, testable agents with no code required.
Customer sectors: software developers, AI product teams, no-code enthusiasts
-
Agent Contracts
— Defines and enforces expected behaviors for AI agents to ensure reliability and accountability.
Customer sectors: AI developers, product managers, quality assurance
-
Continuous Eval
— Continuously tests and evaluates AI agent performance to maintain quality and detect regressions.
Customer sectors: AI engineers, devops teams, quality assurance
Market research & competitive positioning
Source-backed TAM / SAM / SOM sizing, displacement analysis, and an interactive peer positioning map for Relari — available to members.
Related Companies
Other startups in Data Management Platforms
