We value your privacy. We use cookies to enhance your browsing experience, serve personalized ads or content, and analyze our traffic. By clicking "Accept All", you consent to our use of cookies. Read our Privacy Policy for more information.

Building an AI agent to match patients with clinical trials

Healthcare

Product innovation

Cloud

SaaS

Amazon Web Services

Curewiki is a startup on a mission to democratize access to clinical trials for patients worldwide. They're revolutionizing the process by shifting from a research-centric recruitment model to a patient-centric one, empowering patients to find trials that are right for them. The Curewiki team approached us with ambitious goals and entrusted us with a full-service mandate for the end-to-end digital product development.

Challenge

We faced a two-fold challenge: (1) designing & building a seamless end-to-end workflow for both researchers and patients, and (2) developing a system to automatically interpret the medical text within clinical trials, accurately identifying eligibility criteria for patient participation.

Solution

We delivered a comprehensive digital platform with a user-friendly experience for both researchers and patients. Alongside this, we established operational tools and practices for the platform and provided the necessary documentation for data privacy and security compliance.

Approach

We assembled a dedicated product team, comprising a product manager, a designer, and a team of skilled software engineers, working in close collaboration with Curewiki. To build the AI agent, we utilized highly specialized medical large language models (LLMs) to interpret clinical trial content, while simultaneously training the AI agent to gather the relevant patient health information.

Solving a two-sided marketplace problem with AI

To ensure the success of the Curewiki model, we needed to create value for both patients and researchers, while adhering to all regulatory considerations.

Distinct conversational flows for patients/researchers

We built a patient-facing bot, which employs targeted questioning to pre-screen individuals while prioritizing privacy and collecting only relevant data; and a researcher-facing workflow tool designed to validate trial criteria and improve patient communication, thereby addressing the challenges clinical trial managers face in identifying suitable participants.

Highly specialized medical language models

The Curewiki agents are specifically tailored towards the medical industry, with vector embeddings linked to clinical research data and dedicated LLM’s for medical purposes, e.g., Amazon Comprehend Medical.

‍

Organising a highly effective product development process

Co-creation:

We set up a dedicated product development team consisting of 4 software engineers and a fractional product manager together with the domain experts from the Curewiki team.

Release early, release often:

Curewiki operates as a serverless SaaS product built on AWS infrastructure. We implemented automated pipelines with extensive test coverage to enable frequent and reliable releases.

Validating with users:

We developed several tools to interact with users and ensure that their needs are met. User feedback plays a crucial role in shaping our key product roadmap decisions.

‍

It’s remarkable how well Panenco grasped our vision and ambitions. Given that Curewiki was still in its early stages, we needed an entrepreneurial partner that could make the bridge from strategy to product delivery. That’s exactly what Panenco provided to us, in a spirit of true co-creation. - Jean-Sebastien Gosuin, Co-founder & CEO Curewiki

The Curewiki team is tackling an enormous global problem that only recently became feasible to solve through the advancements in the field of generative AI. It’s heartwarming for us to consider how many lives this technology will touch.

We wish the Curewiki team the very best of luck as they are driving the go-to-market of this technology. We’ll gladly keep supporting you every step of the way!