Anthropic Claude + AWS: revolutionizing pharma data analytics with Snorkel AI
A leading pharmaceutical company has committed to double its revenue by 2030 and aims to fuel that growth, in part, with AI-powered data insights.
Seeking to build an AI system that could extract, analyze, and present insights from vast, complex datasets, the company partnered with Snorkel AI, Amazon Web Services (AWS), and Anthropic. The company sought the trustworthy results of Anthropic’s Claude models, the security and cost controls of Amazon Bedrock, and the ability to rapidly, expertly, and reliably curate training data provided by the Snorkel AI Data Platform—which integrates natively with AWS.
Combining these tools, the firm created an agentic AI co-pilot capable of better navigating its data, unlocking critical business insights, and driving decision-makers’ ability to identify opportunities and challenges across its operations.
Key Outcomes:
- AI-ready data: Snorkel’s programmatic approach to data curation—labeling, sampling, filtering, and augmenting data—helps AI teams efficiently capture expert knowledge to build high quality datasets and iteratively build production-quality AI.
- Accelerated AI development: Snorkel’s integration with Amazon SageMaker and Amazon Bedrock enabled rapid fine-tuning and deployment.
- Enhanced data understanding: The resulting application, built on Anthropic’s Claude Sonnet model, empowered key decision-makers with up-to-date insights without requiring them to know how to code.
This collaboration demonstrates how integrated AI solutions can effectively address data complexity challenges and set a new standard for AI adoption in the pharmaceutical industry. It created tangible improvements in operational efficiency and business performance, supporting long-term growth objectives.
AWS and Amazon Bedrock
Amazon Bedrock offers a fully managed service that provides seamless access to leading foundation models, including Anthropic’s Claude series. This integration facilitates the development and deployment of generative AI (GenAI) applications without extensive setup or specialized infrastructure.
Advantages of AWS:
- Scalability and performance: Bedrock’s robust infrastructure ensures that enterprises can scale their AI applications efficiently.
- Cost-effective AI solutions: AWS’s managed services allow enterprises to optimize costs associated with deploying and maintaining AI applications.
- Efficient fine-tuning: Snorkel integrates with SageMaker and Bedrock to orchestrate streamlined model fine-tuning to optimize AI performance.
The pharmaceutical giant chose Amazon Bedrock for several reasons. Bedrock provides a comprehensive, secure, and efficient platform for enterprises. It integrates seamlessly with Snorkel’s ai data development platform and allows companies to access and deploy Anthropic’s Claude models, which are aligned with their organizational goals of innovation, safety, and operational excellence.
Anthropic Claude integration
Anthropic’s Claude models, accessible via Amazon Bedrock, offer:
- Advanced reasoning: Claude handles complex problem-solving tasks by integrating diverse data points for coherent conclusions.
- Multimodal analysis: It interprets and analyzes visual data alongside text.
- Code generation: Claude facilitates code creation using natural language descriptions, which helps generate queries for the pharmaceutical giant’s data management systems.
- Multilingual support: Enables effective communication across global teams by supporting multiple languages.
Anthropic’s Constitutional AI approach underpins the Claude models with a principled framework aligned with human values. This reduces the risk of harmful or biased outputs—enhancing trust, reliability, and transparency.
Claude Models overview
Anthropic offers three Claude models tailored to specific use cases:
- Claude Opus 4: Anthropic’s largest hybrid reasoning model, Opus excels in complex tasks requiring high accuracy and advanced language comprehension.
- Claude Sonnet 4: Balances capability and performance, making it suitable for general business applications such as coding assistance and enterprise deployments.
- Claude Haiku 3.5: Haiku is optimized for speed, cost-effectiveness, and agentic tool use, making it ideal for applications like customer support and content moderation.
Snorkel AI Data Platform capabilities
Snorkel’s AI data development platform accelerates the process of converting raw records into high-quality training data sets by 10-100x by combining:
- Programmatic data curation: Experts contribute logic that data scientists encode into labeling functions. The platform applies these labeling functions to the entire dataset, using Snorkel’s proprietary weak supervision algorithm to apply the most likely label when they conflict. This minimizes the manual effort required to build training data while improving label consistency and auditability.
- Guided error analysis: Snorkel’s guided error analysis helps users identify shortcomings within the training data for targeted improvement, facilitating iterative refinement.
- Integration with enterprise infrastructure: Snorkel integrates seamlessly with enterprise cloud infrastructure, including AWS, ensuring scalability and security.
- On-board annotation suite: Snorkel’s integrated annotation suite enables SMEs to manually create additional labels where needed.
Snorkel’s tools and features empower enterprise data science teams to iteratively improve models until they reach production benchmarks—meeting the challenges of the pharmaceutical industry and many others.
Putting it all together
The pharmaceutical giant aimed to build an advanced AI system that could effectively query, visualize, and explain data accessible through its existing database tools and APIs. However, the team faced significant challenges, including a slow user acceptance testing (UAT) process. This hindered the collection of organic training data and slowed progress.
To overcome these challenges, Snorkel researchers collaborated with the pharmaceutical company to develop a process using Anthropic’s Claude models to programmatically generate, filter, curate, and evaluate synthetic UAT data. Additionally, Snorkel’s researchers helped distill a smaller guardrail model that could be cost-effectively deployed on Amazon Bedrock, ensuring robust pre- and post-production reporting and flagging potential errors in AI outputs.
Better together: Snorkel AI, Amazon Bedrock, and Anthropic Claude
This partnership represents a paradigm shift in how AI collaborations can drive business transformation.
By combining Anthropic’s cutting-edge language models, Amazon Bedrock’s enterprise-grade deployment capabilities, and Snorkel AI’s powerful AI data development platform, the pharmaceutical company created an AI system that empowers decision-makers with rapid insights.
Key collaborative benefits:
The partnership between Snorkel AI, AWS, and Anthropic yielded significant benefits for the pharmaceutical company, transforming its AI capabilities and operational efficiency.
- Time savings: The company accelerated AI development by automating data labeling through Snorkel’s tools and leveraging synthetic data from Anthropic’s Claude models. This reduced manual annotation time and enabled faster deployment without compromising quality or compliance.
- Accuracy improvements: Snorkel’s labeling functions and weak supervision enhanced model accuracy by ensuring high-quality training data.
- Cost reductions: The partnership optimized costs by leveraging Amazon Bedrock’s scalable infrastructure.
This collaboration demonstrated how integrated AI solutions can effectively address AI challenges in the pharmaceutical industry, improving operational efficiency, model accuracy, and cost management.
Learn more about Snorkel + AWS
Snorkel AI and AWS unlock the power of AI by empowering some of the world’s leading companies to transform their data and knowledge into real-world business value. Snorkel’s platform is also available through the AWS Marketplace. If you would like to learn more about what Snorkel can do for your organization, book a demo today.
Shan is a Senior Partner Solutions Architect specializing in Generative AI at AWS, dedicated to solving complex customer challenges. He advocates for innovative AI solutions, distributed architecture, and serverless technologies, helping customers harness the power of Generative AI in their cloud journey.