Biotech Firm Pluton Biosciences Eliminates AI Hallucinations in Microbial Research with AWS Bedrock
At a glance
Pluton Biosciences, a computational biology firm developing microbes for agriculture and climate solutions, struggled with LLM inaccuracies and hallucinations that made AI unreliable for research acceleration.
NMD deployed a RAG system using AWS Bedrock with Amazon Titan and Anthropic’s Claude that eliminated hallucinations through structured, cited outputs and automated data pipelines integrating lab and scholarly data sources.
The solution scans hundreds of papers per microbe and returns results within 10 minutes at $0.12 per taxonomic analysis. An image processing POC is now underway to automate biological sample analysis.
Industry
Use Case
- Natural Language Processing
- Knowledge Management
- Analytics & Business Intelligence
Solution implemented
- AWS Bedrock RAG architecture with Amazon Titan and Anthropic's Claude models
- Lambda-based fanned-out processing scaling with sample volume
- Automated pipeline vectorizing lab data and scholarly databases
- Terraform infrastructure-as-code for reproducible deployments
- [POC 2] Image processing pipeline with well extraction and NDVI vegetation index calculation
The value equation
- Automated analysis scanning hundreds of papers per microbe with 10-minute turnaround
- Eliminated AI hallucinations via structured, cited output format
- Scalable Lambda architecture handles growing sample volumes without redesign
- $0.12 per taxonomic name analysis enabling cost-effective research workflows
- Foundation for automated image processing and biological data extraction
Company Snapshot
Pluton Biosciences discovers and analyzes novel microbes for agriculture and climate solutions through computational biology, bringing together scientists and technologists to protect the planet.
Location
St. Louis, Missouri, USA
Customer Situation
Pluton Biosciences faced critical challenges preventing AI adoption in microbial research. LLMs produced inaccurate results and hallucinations, making them unreliable for research acceleration. Their biological data from lab samples and scholarly sources existed in unstructured formats that couldn’t be transformed into actionable insights.
In biotechnology, computational biology firms must analyze vast amounts of unstructured data from experiments and literature, but off-the-shelf LLMs lack domain expertise and frequently generate false information.
NMD Solution
NMD found the core challenge was the disconnect between LLMs’ linguistic abilities and essential domain knowledge, requiring a RAG architecture that augments AI with precisely relevant scientific data.
Enabling accurate AI-driven research required implementing a RAG system using AWS Bedrock for LLM orchestration, Lambda for scalable processing, and Terraform for infrastructure, with structured output formats ensuring cited, verifiable results.
What We Delivered
Within 8 weeks of deploying NMD’s RAG-based Knowledge Management solution, Pluton Biosciences eliminated AI hallucinations and achieved automated analysis that scans hundreds of papers per microbe and returns highly accurate, cited results within 10 minutes at $0.12 per taxonomic name analysis. The Lambda architecture scales automatically without redesign, processing both lab-generated and scholarly data sources.
The success led to POC #2: an image processing solution automating biological sample analysis including well extraction, shape isolation, and NDVI calculations.
Ready to Transform Your Biotechnology Business?
See how New Math Data can transform your biotechnology business with AWS-powered innovation.