Cradle AI

Cradle AI has emerged as a transformative force in protein engineering, leveraging advanced AI algorithms to accelerate the development of therapeutic, agricultural, and industrial proteins. This article explores the company's technological platform, its scientific achievements, and its impact on the biotech industry, while highlighting the rigorous data privacy protocols that underpin its operations.

Company Overview and Funding

In November 2022, the company raised $24 million in Series A funding, led by Index Ventures with participation from Kindred Capital, Recursion co-founder Chris Gibson, and Thomson Reuters former CEO Tom Glocer. This initial investment brought their total funding to $33 million.

The company experienced substantial growth following its Series A, signing partnerships with major players including Janssen Research & Development, Novozymes, and Twist Bioscience. Since then, they have onboarded 12 additional research and development projects across diverse protein modalities such as enzymes, vaccines, peptides, and antibodies.

In July 2024, Cradle completed its Series B funding round, raising an additional $73 million. This brings their total funding to over $100 million, with IVP leading the investment and continued support from Index Ventures and Kindred Capital.

The company employs 40 individuals across offices in Delft and Zurich, with a focus on combining wet lab research with AI capabilities. Their team structure includes 25 full-time employees (FTE), with expansion plans underway for their Amsterdam office.

Technology and Operations

Cradle's platform integrates machine learning algorithms that enable users to design improved protein variants through a user-friendly interface. The system supports both lab data import from ongoing projects and starting from single sequences, providing immediate context-aware predictions.

The core technology operates through a series of automated workflows that combine sequence generation, scoring, downselection, and active learning criteria. Key technical components include homologous sequence retrieval and alignment using mmseqs2, base model training on evolutionary context, and ensemble fine-tuning through ranking losses.

The platform demonstrates exceptional performance when fine-tuned for specific applications, achieving state-of-the-art results on multiple benchmark challenges. In one particularly challenging enzyme optimization task, Cradle's models outperformed all competitors, demonstrating superior ability to rank protein sequences accurately - a critical metric for effective lead optimization.

Customer feedback indicates significant time and cost reductions, with reported improvements ranging from 1.5x to 12x in research efficiency across multiple project modalities. This performance improvement is compounded through round-over-round active learning, allowing each experimental iteration to build upon increasingly accurate predictions.

Team and Culture

Cradle AI employs 40 individuals across its offices in Amsterdam and Zurich, with a team of 25 full-time employees (FTE). The company requires team members to work at least two days per week in their offices and labs, offering remote work opportunities for other days. Current positions range from bioinformaticians and protein scientists to scientific consultants and software engineers, with recent expansion plans for their Amsterdam office.

The team structure combines expertise from leading tech and biotech companies, including Google, Uber, and Perfect Day, in both Amsterdam and Zurich. Notable team additions include Sam Partovi as Chief Commercial Officer, bringing valuable experience in scaling life science platforms. The company provides comprehensive support for non-EU/EEA or Swiss residents, including relocation assistance and visa sponsorship for their office in the Netherlands.

Cradle emphasizes a collaborative approach between humans and AI, allowing significant team autonomy while maintaining strong connections between Switzerland and the Netherlands through quarterly team meetings. The company offers an extensive benefits package, including 25+ days annual leave, 4 months parental leave, and comprehensive training opportunities. Salaries are competitive, with employees receiving generous equity stakes in the company.

The work environment prioritizes employee well-being, with part-time work options available and flexible work hours and location arrangements. However, all team members must conduct wet-lab experiments in the company's facilities to maintain the necessary data isolation protocols. The team structure supports rapid growth, with plans to expand the engineering team to tackle more complex challenges and build additional laboratory facilities in Amsterdam.

Applications and Partnerships

Cradle's technology has been commercialized with 21 customers working on 31 proteins, including partnerships with Janssen Research & Development, Novozymes, and Twist Bioscience. The company's platform enables scientists to design improved protein variants through a user-friendly interface, combining both wet lab data import from ongoing projects and starting from single sequences to provide immediate context-aware predictions.

The platform has demonstrated significant performance improvements when fine-tuned for specific applications, achieving state-of-the-art results on multiple benchmark challenges. One particularly successful application was in enzyme optimization, where Cradle's models outperformed all competitors by demonstrating superior ability to rank protein sequences accurately - a critical metric for effective lead optimization.

Financial benchmarks indicate that the technology enables most projects to progress two times faster than industry standards, with reported improvements ranging from 1.5x to 12x in research efficiency across multiple project modalities. The platform also reduces development costs by up to 90% compared to traditional methods, with customer satisfaction scores averaging 8+ across all projects.

Ethics and Data Privacy

Cradle AI employs rigorous data privacy protocols to maintain the confidentiality of model weights derived from customer data, treating them with the same care as the underlying data itself. The company achieves this through multiple layers of security barriers that completely isolate model weights for each customer account. While the machine learning team continuously benchmarks algorithm performance across various datasets to improve model accuracy, this process never involves sharing or transferring specific data or model weights between customers.

The benchmarking and quality control framework operates on public and private datasets, with strict permissions requirements. Once benchmarking sessions conclude, all intermediate models are explicitly destroyed, preventing any residual data retention. This approach enables the company to stay at the forefront of AI development while maintaining absolute data privacy standards.

Recent technical advancements have yielded impressive results in protein engineering accuracy. During the 'Align to Innovate' benchmark, Cradle's models demonstrated exceptional performance across four enzyme families, achieving state-of-the-art results while requiring fewer computational resources than traditional methods. The company's platform combines predictors and generators to evaluate and propose protein variants, with generator performance serving as a strong indicator of real-world effectiveness.

The platform's automated workflows have processed multiple proteins across diverse applications, including therapeutic, agricultural, and foodtech domains. Current customers report efficiency improvements ranging from 1.5x to 12x across their projects, while several partnerships have achieved significant milestones, such as an 8x improvement in EGFR binding affinity through the Adaptyv Bio protein design competition.