AI for R&D
Helping researchers build on what their team already knows.
A company that does a lot of research also collects a lot of knowledge. Reports from earlier experiments, measurements, scientific papers, and observations that might be useful again later. But when a new question comes up, finding the right piece of information can be difficult. Has someone tried this before? What happened, and under which conditions?
At Hammer, I worked on an AI-supported R&D platform in collaboration with a research team in the food industry. The idea is to make their existing knowledge easier to use, and connect it with the scientific literature they work with.
My work connects document processing with retrieval-augmented generation (RAG). Research reports contain more than paragraphs of text: results can sit in tables, charts or scanned pages. We process these documents so that researchers can ask questions across the collection and follow the references back to the original evidence.
This also makes it possible to explore new ideas. A finding in a scientific paper might help explain an internal experiment, or suggest something worth testing next. The platform helps researchers work through these connections, with a distinction between what the documents establish and what is still a hypothesis.
Beyond asking questions, the platform suggests relevant papers and reviews proposed experiments against the available literature. It can highlight supporting findings, contradictions and gaps before a team commits time in the lab. An initial patent screening also flags possible overlaps for further review by a specialist.
For me, the interesting part is bringing AI into an actual research workflow. A useful answer needs to give the researcher enough context to judge it and decide what to do next.
You can try the interactive demo to see how this works. It uses fictional documents and prepared examples, so none of the client's research is shown.