Cambrium Advances Protein Design and Develops Sustainable Materials 5 Times Faster Using AWS

Founded in 2020, Cambrium set out to transform the materials industry. “Materials are everywhere around us,” says Pierre Salvy, chief technology officer of Cambrium. “They’re also responsible for 23 percent of global greenhouse gas emissions.” To develop more sustainable alternatives, Cambrium requires a robust infrastructure that can handle complex protein design and analysis at scale. In early 2024, the startup migrated its protein vectorization engine to AWS to scale its protein design pipeline. Cambrium also participated in the NVIDIA Inception, program, gaining early access to deep-learning containers that are optimized for life sciences workloads.

In just one month, a single Cambrium team member can build an entirely new protein analysis pipeline on AWS. When new protein files are uploaded to Amazon Simple Storage Service (Amazon S3), an object storage service, the pipeline is automatically triggered. This initiates a job running on AWS Batch, a fully managed batch computing service, to prepare and clean the data.

These prepared files are then processed using NVIDIA-accelerated containers, which transform the proteins into vectors that are stored for future analysis. The company uses NVIDIA’s BioNeMo, a generative AI platform for chemistry and biology, for this. Using BioNeMo–provided ESM models makes it possible for Cambrium to seamlessly integrate optimized protein vectorization endpoints into the stack.

The startup also uses Amazon SageMaker, a fully managed machine learning service, to manage its machine learning operations and develop new AI models. Using Amazon SageMaker, Cambrium develops and trains its own foundation models for protein analysis.

“The real value add of AWS is scalability,” says Salvy. “We can ramp up our test dataset from 5 items to 5,000 toa million within 2 hours.” Using this pipeline, Cambrium can process queries across millions of proteins in less than 100 milliseconds.

With these capabilities, Cambrium can develop sustainable materials 5 times faster and 20 times more cost-effectively than with traditional industry methods. The startup has used its pipeline to develop NovaColl, a sustainable collagen created through precision fermentation rather than animal products. This bioidentical protein not only produces 95 percent fewer emissions than traditional collagen but also outperforms it in skin care applications.

But this is just the beginning. “With the other materials that we have planned, we have the potential to save 4,000 million tons of carbon emissions by 2050,” says Salvy.

Given Cambrium’s ambitious future plans, using the right technology is crucial—and the startup will continue to innovate on AWS. “The scalability and integrated AI services of AWS are exactly what we need,” says Salvy.