Benjamin Fletcher is chief technology officer of Valantor and co-founder of EyeLevel.ai, creator of GroundX, an enterprise platform for extracting reliable answers from complex documents. His work targets a persistent weakness in retrieval-augmented generation: even sophisticated language models fail when their source material is extracted incorrectly, separated from essential context, or retrieved inaccurately.
Fletcher holds a doctorate in physics and previously worked at IBM Research on quantum-computing chips and mobile security. At IBM Watson, he led AI engineering for Watson Life and Watson Weather and helped develop The Weather Channel’s conversational platform. His earlier career also included roles at The Weather Company and Weather Underground.
He co-founded EyeLevel in 2019, serving as chief executive and chief scientist as the company developed GroundX and FraudX. Valantor acquired EyeLevel in July 2026, bringing GroundX into its enterprise visual-intelligence platform and making Fletcher its chief technology officer.
- Document ingestion determines answer quality. Fletcher traces many enterprise AI failures to upstream processing: garbled text, discarded tables and images, and document fragments stripped of their original surroundings. His Air France customer-support example demonstrates how ordinary OCR can erase information essential to interpreting airline policies.
- Semantic objects preserve document context. His retrieval architecture combines source passages, generated metadata, and separate representations optimized for search and answer generation. Query rewriting, multifield search, fine-tuned vision models, multimodal extraction, and language-model reranking recover relationships that simple chunking misses.
- Vector retrieval changes at enterprise scale. Fletcher collaborated with EyeLevel engineer Daniel Warfield on research into retrieval accuracy as document collections grow, challenging assumptions about vector-search reliability through company-reported findings.
- Private enterprise AI should run where the data lives. GroundX supports private-cloud, on-premises, sovereign, and air-gapped deployment architectures for organizations that cannot transfer sensitive documents to external services.
