Jonathan Lowe is an enterprise data-science leader specializing in knowledge graphs for pharmaceutical manufacturing. His work makes scientific records, supply-chain data, and operational expertise accessible to the teams responsible for turning laboratory discoveries into medicines produced at industrial scale.
Lowe’s career includes positions at Informix, IBM, Deloitte, Boston Consulting Group, and Pfizer, alongside founding companies in Berkeley and London. Early work modeling agricultural subsidy rules sharpened his interest in representing complicated real-world relationships as structured data. At Pfizer, he served as a senior director and data-science lead for Operations & Insights, applying graph databases and analytics to manufacturing and supply chains.
His projects included vaccine inventory and supply planning and commercial organic synthesis planning, for which his Digital Insights team helped implement a graph database and user interfaces. He has also emphasized building teams that combine data science, software development, consulting, and manufacturing expertise.
- Biopharmaceutical technology transfer: Scaling drug production requires manufacturing teams to absorb experimental results, scientific documentation, and expertise dispersed across departments. Lowe applies generative AI and connected data to preserve that institutional knowledge and make it usable during the transition from development to production.
- Graph-based document chunking: Lowe organizes documents, blocks, paragraphs, and lines as connected graph elements, allowing teams to test which level retrieves the strongest evidence. Connected neighboring information supplies context that isolated similarity matches can overlook. His AI Engineer Summit conversation also describes how shared graph structures help technical teams understand unfamiliar organizational data.
- Natural-language access to structured data: Lowe built a local knowledge-graph prototype combining Neo4j, Python, sentence embeddings, semantic retrieval, graph traversal, and locally run language models. It converts graph relationships into searchable text, retrieves relevant connections, and lets nonspecialists interrogate structured information without writing database queries—a goal he also develops in his professional writing.
- Enterprise adoption through measurable value: Lowe designs internal AI products around accurate answers, responsive interfaces, executive priorities, departmental ownership, and existing systems. He argues that technical teams must explain proposals in operational terms decision-makers can evaluate: cost, timing, revenue, staffing, and concrete business outcomes.
