Overview

As a pioneer in allogenic cell therapies, the Company manages enormous volumes of research documentation. From peer-reviewed publications and preclinical trial data to internal experimental reports, the sheer weight of information often slowed the ability of scientists, clinicians, and executives to extract the insights that mattered most.

The problem was not access — the Company had plenty of data — but speed and clarity. Key findings were often buried in 50-page reports or technical papers that took hours to digest. Scientists were spending precious time parsing documents instead of running experiments. Executives, meanwhile, needed concise and regulator-ready summaries to make informed strategic decisions.

The Company recognized the potential for Generative AI (GenAI) to transform this workflow. However, like many biotech companies working with sensitive data, the Company had strict security requirements that ruled out SaaS-based LLMs. Public AI services carried risks of intellectual property leakage and compliance violations. The Company turned to PTP to design a secure, AWS-native summarization solution that met both technical and regulatory needs.

The Challenge

The barriers the Company faced were familiar across biotech, but particularly acute in high-stakes cell therapy research:

Overview

A research-driven biotech is advancing its pipeline through data-intensive drug discovery and clinical development. Among the most resource-heavy steps in this journey is clinical trial design—a process requiring teams to comb through thousands of historical studies, extract eligibility criteria and endpoints, and draft complex protocols that meet regulatory standards.

While critical to bringing new therapies to patients, protocol design is time-consuming, repetitive, and a frequent bottleneck. The Company sought to test whether Generative AI (GenAI) agents built on AWS Bedrock could streamline trial design, accelerate protocol drafting, and improve consistency across its development programs. Partnering with PTP, the Company launched a proof of concept (POC) centered on two Bedrock-powered clinical development agents, laying the foundation for an extensible GenAI framework to support future R&D needs.

The Challenge

Designing and validating clinical trial protocols introduced two major challenges for The Company:

These inefficiencies slowed R&D progress, delayed hypothesis testing, and consumed valuable researcher time. The Company’s goal was clear: use GenAI to automate repetitive tasks, generate consistent protocol drafts, and free its scientists to focus on innovation—all while staying within compliance boundaries by using public, non-sensitive data.

Overview

A clinical-stage biotechnology company, focused on engineering next-generation proteins to accelerate therapeutic innovation, was searching for AI-enabled advancements to their research. At the heart of their pipeline were machine learning (ML) models that predicted protein folding and interaction patterns, helping researchers identify promising therapeutic candidates. While these ML models delivered powerful predictive capabilities, the company’s scientists faced a persistent bottleneck: turning raw predictions into actionable insights.

Protein research is inherently interdisciplinary, requiring collaboration among computational biologists, molecular modelers, chemists, and wet-lab researchers. While ML systems such as AlphaFold could produce detailed folding predictions, these outputs often needed extensive interpretation and translation into experimental briefs. This process consumed valuable time and slowed experimental cycles, hindering the company’s ability to quickly iterate and validate new therapeutic hypotheses.

To address this challenge, the company partnered with PTP to integrate its existing ML pipeline with Generative AI (GenAI) capabilities on AWS Bedrock. The result was a transformative workflow that combined the predictive power of ML with the contextualization strengths of GenAI. Predictions became clear, plain-language, experiment-ready briefs that allowed interdisciplinary teams to collaborate more effectively, shorten research cycles, and accelerate the development of new protein-based therapeutics.

The Challenge

The company’s research bottlenecks were shaped by three interrelated challenges:

Interpretation Gap

The company’s ML models could generate folding predictions and structural interactions, but these outputs were dense, technical, and difficult for non-specialists to interpret quickly. Cross-functional teams had to spend significant time translating computational predictions into insights usable for experimental design.

Time-Consuming Summarization

Reports summarizing ML outputs were drafted manually by data scientists and computational biologists. Each cycle required days of analysis and writing, extending experimental planning cycles and delaying downstream work.

Scaling Research Output

As the company expanded its protein engineering pipeline, the number of candidate proteins under investigation grew dramatically. Scaling human effort to match ML output was not feasible, creating a widening gap between computational predictions and actionable experimentation.

The company set a clear goal: Join ML to GenAI in a seamless pipeline that could automatically generate structured, comprehensible, and actionable reports—without sacrificing scientific rigor or compliance.