Amazon Web Services unveiled Amazon Bio Discovery, an AI platform that provides researchers with a library of biological foundation models that can be fine-tuned for drug discovery, supported by an AI agent that selects models and evaluates synthesis candidates for a rapid lab-in-the-loop cycle.

Lab partners test the most promising antibody candidates and feed results back to scientists, creating a feedback loop that refines subsequent design iterations.

The platform operates on a subscription model with a five-experiment free trial and deliberately avoids ads in search results.

Collaborative workflows encode decisions on model chaining, data processing, and quality thresholds, enabling validated results to continually improve models and shorten queue times.

It enables parallel experimentation and automates parts of the analysis to reduce manual intervention.

Results flow back into the platform to compare predictions with actual outcomes, driving iterative refinement of workflows.

Analysts note that queued projects can move forward immediately and computational biologists can support more programs with improved workflows.

Computational biologists can build reusable, no-code workflows while bench scientists run multiple experiment variants with agent-guided inputs, boosting collaboration and speed.

AWS’s Luca Giancardo underscored the Developability Benchmark as a key measure for evaluating model suitability and performance.

The platform emphasizes data isolation and ownership, enabling researchers to train custom models on their own data with private fine-tuning and in-house deployment options.

Researchers can train models on their data to improve predictions while maintaining privacy and ownership.

A companion article highlights AI’s transformative potential for drug testing workflows, though technical specifics aren’t detailed in the text.