As a team of former bench scientists, we experienced challenges finding commercial antibodies. So we spent two years building machine learning software to extract antibody usage data in the form of published figures, decoding millions of papers and making the data easily discoverable for scientists. The result: BenchSci. For biomedical researchers who are starting experiments, BenchSci is a reagent intelligence platform that transforms published data into experiment-specific recommendations to reduce time, money and uncertainty in planning materials and methods. Unlike PubMed, Google Scholar, reagent directories and vendors, BenchSci uses machine learning to decode open- and closed-access data and present published figures with actionable insights.