Jack Boylan

At Quantexa, I build, train, deploy, and operate production NLP models for structured information extraction, including named entity recognition, coreference resolution, relation extraction, and document classification.

I turn customer-reported extraction failures into labelled evaluation cases and model releases, while developing hardware-scalable inference pipelines for enterprise workloads under tight compute constraints. I also lead NLP research, shape product direction through work published at NAACL, ACL, COLM, and ESWC, and define practical LLM approaches for enterprise unstructured data.

Previously, I built PySpark pipelines for recommendation models at VARTEQ and deployed real-time polyphonic music transcription at Fretello, achieving under-15 ms inference latency across edge devices. At Language Weaver, I built neural machine translation models across seven-plus languages and helped reduce translation errors by 61%.

I'm the creator and maintainer of GLiREL, a CPU-friendly zero-shot relation extraction model with 290+ GitHub stars, and an open-source contributor to grand-cypher for NetworkX graphs. My work spans low-resource extraction, multilingual NLP, synthetic data, knowledge graph validation, and graph-based machine learning.