AI/ML Model Development for Protein Aggregate Classification

The contract seeks the development and training of advanced machine learning models to identify, classify, and analyze subvisible protein aggregates in biosimilar drug products using morphological fingerprinting and neural network-based metric learning, all without relying on traditional image pre-processing techniques. The goal is to enable accurate, automated detection of protein aggregates through direct feature extraction and pattern recognition from raw data, improving the reliability and efficiency of quality control in biopharmaceutical manufacturing. The approach emphasizes end-to-end learning where the model learns discriminative representations directly from unprocessed inputs, reducing artifacts and preserving biological relevance. The work is part of a subcontract under the FDA Office of Acquisition and Grant Services, with performance to occur in Silver Spring, Maryland, at the zip code 20993. The solicitation was posted on May 11, 2026, and responses are due by May 26, 2026, at 5:00 PM. The North American Industry Classification System code 541715 indicates this is for research and development in the physical, engineering, and life sciences, specifically targeting AI and machine learning applications in pharmaceutical analysis. The project is focused on advancing automated analytical capabilities for biosimilars, supporting regulatory oversight and product integrity without manual intervention or data preprocessing steps.