AI assistant cuts compound research from months to hours

The client

Merck is a German science and technology group active in healthcare, life science and electronics, employing around 62,000 people in 65 countries. Founded in 1668, it is the oldest operating chemical and pharmaceutical company in the world, and its experts continuously monitor scientific publications for new chemical compounds.

The challenge

Domain experts spent around six months manually combing literature to identify relevant molecules. The volume of text exceeded what mainstream LLMs could take in at once, so documents had to be split before processing. Merck's internal catalogue also lacked identifiers such as InChIKeys and CAS numbers, which blocked automatic matching against products. On top of that, the proof of concept had a hard five-week deadline and a fixed budget.

The solution

The results

The AI R&D Assistant reduced a roughly six-month manual research task to about six hours. The proof of concept shipped inside the five-week window and under budget, hosted on Merck's own AWS infrastructure. Because Merck operates in 65 countries, the tool is relevant across its global business.

“If I had to choose two words to summarize our collaboration with Netguru it would be speed and efficiency.”

Tools & stack

Original project by Netguru ↗ — published by the delivering agency, included here as an industry reference.