Machine learning (ML) offers a transformative solution by enabling predictive modeling of process behavior and optimal conditions. However, conventional ML approaches typically require large datasets (i.e. 100’s-1000’s of points), which are expensive and impractical to generate in many R&D environments. This limitation is addressed through Bayesian Optimization frameworks, which guide experimental campaigns by selecting only the most informative experiments—making accurate modeling possible with dramatically fewer data points.
At Sunthetics, we’ve built a modified Bayesian Optimization approach tailored for low-data environments. In one foundational study, we demonstrated that reliable modeling can begin with as few as five data points. While initially applied to an organic electrosynthetic reaction, this approach has since scaled to diverse applications across pharmaceutical process development—including biocatalysis, biomass conversion, dehydrogenation, and time-series modeling for cell cultures. We continue to expand its capabilities by incorporating flexible constraints and deterministic outputs to reflect real-world manufacturing conditions.
SuntheticsML, our web-based platform, empowers scientists and engineers to optimize chemical processes, formulations, and materials—no ML or coding experience required. The platform features proprietary algorithms designed for small-data ML, accelerating development timelines by up to 32x. Recent enhancements include double-sided constraints, categorical variable support, time-dependent modeling, and automated reporting for deeper process insights.
This work helps redefine what’s possible in pharmaceutical process development by dramatically reducing the number of required experiments, improving decision-making, and supporting more sustainable, efficient innovation. SuntheticsML is lowering the barrier to ML adoption across the chemical industry—enabling smarter, faster, and greener R&D.