Treatment Side-Effect Forecasting AI

With a healthcare business, we developed AI that forecasts side effects arising from treatment.

What we built

Two technologies. The first is a multivariate time-series model that predicts the risk of side effects from longitudinal data such as medication records and lab values. The second is a small language model (SLM) that answers questions with specialist knowledge about side effects. Both run in a local environment so that medical data never leaves the premises.

Outcome

The system combines answer quality comparable to large models with fast responses, entirely on local infrastructure. It catches early signs of side effects and supports response decisions, in a form designed for clinical use.