Formulation development cycle down 30%, duplicate experiments down 40%

Twenty years of lab records and formulation data, tens of thousands of them, sat on researchers' laptops and in paper notebooks, and new projects routinely repeated experiments that had already been done. When senior researchers left, their experience left with them, and a new hire took three years to develop a formulation independently.

The company wanted R&D experience captured as an organizational asset, and faster new-product response to customer requirements.

Starting from experimental data governance, we launched an R&D knowledge Agent to support formulation design and experiment planning.

• Historical lab records structured: formulations, process conditions and test results consolidated into one experiment database
• Formulation assistant: retrieves similar past formulations for a performance target and proposes a starting formulation and adjustment directions
• Experiment planning: recommends a design of experiments that avoids points already tested
• Every completed experiment enters the database, so the knowledge base grows in step with R&D

Formulation data is a core trade secret, so everything is processed in an on-premises environment with a privately deployed model. Similar-formulation retrieval uses ingredient and performance vectors, with access isolated by project team.