Atla builds software for testing and improving AI agents by identifying failure patterns across runs. Its system uses an LLM judge to evaluate agents step by step, then suggests specific fixes for recurring errors. The company also trains open-source judges called Selene and Selene Mini, which have been downloaded more than 60,000 times.
Co-founder & CEO
Co-founder & CTO
York House, Fora, 221 Pentonville Rd, London N1 9UZ, United Kingdom
Atla primarily focuses on the artificial intelligence (AI) industry, specifically in providing evaluation models for generative AI applications to help businesses assess their reliability and performance.