Helping Tachmed bring at-home testing to life.

The Client

Tachmed is redefining healthcare diagnostics with their innovative at-home testing solutions. Leveraging cutting-edge technology, their systems empower patients to conduct accurate, real-time health tests from the comfort of their homes. These devices seamlessly connect via the Tachmed cloud-based platform with healthcare agencies, payers and providers, delivering instant results and actionable insights.

Tachmed’s mission goes beyond diagnostics—it’s about democratizing healthcare, making it more accessible, efficient, and data-driven to address global health challenges and improve patient outcomes. Their approach combines advanced diagnostic tools with responsible and compliant big data collection, ensuring a future where healthcare is proactive, precise, and universally accessible.

The Challenge

To develop cutting-edge diagnostics and bring them to market, Tachmed relies on rapid discovery and evaluation of potential clinical use cases and diagnostic reagents, such as antibodies,  for assay (test) development. However, manual processes for sifting through scientific literature and vendor data are time-consuming, costly, and prone to bottlenecks. Those tasked with doing this are talented scientists whose time is valuable.

This inefficiency is not only expensive but delays development  of Tachmed's expansive assay library, necessary for making the system viable for remote patient use and professional clinical decision-making. Tachmed needed to reduce the amount of time Tachmed scientists spend on developing each new diagnostic, including time spent on identifying clinical use cases, as well as identifying appropriate reagents for each use case.

Our Approach

Daemon worked closely with Tachmed to identify the most impactful opportunities for automation and implementation of an AI-powered solution. The collaboration started with a discovery phase to align with Tachmed’s mission and culminated in a Proof of Concept (POC) application that has been adopted by scientist users and is now actively accelerating their research and development workflows.

• Discovery Phase:
  • In-depth workshops with Tachmed to explore potential use cases for AI and select the highest-value target. Daemon worked with Tachmed scientists to understand the viability and potential value of transcriptomic and statistical foundation and ML (machine learning) models, settling on knowledge distillation for its relatively low barrier to entry and expected efficiencies for saving scientist time.
  • Educational enablement for scientists, developers and non-technical stakeholders on AI’s capabilities, inspiring the vision for a scalable, long-term solution.
• AI-Driven Automation:
  • Use of advanced AI foundation models. Using the highly performant Claude Sonnet 3.5 model from Anthropic we were able to obtain great results very quickly and with little effort.
  • Design of a two-phase tool for distilling knowledge from PubMed and vendor websites:
    • Clinical Uses View: Automates the identification of biomarkers, clinical gaps, and diagnostic opportunities.
    • Reagents View: Aggregates antibody data to streamline selection and decision-making for assay development.
  • Implementation of a dual-phase AI workflow:
    • Extraction Phase: Using AWS Bedrock to extract key data, such as biomarkers, test types, and clinical significance.
    • Consolidation Phase: Again using Bedrock to merge extracted data for streamlined navigation and decision-making.
• Scalable Architecture:
  • Deployment of the application on AWS AppRunner for easy, scalable hosting.
  • Use of Elasticache for fast in-memory operations and RDS Postgres to support future persistence and scalability needs.
  • An asynchronous process flow for handling potentially long-running processes while keeping user experience snappy.
• User-Focused Design and Iteration:
  • Design and implementation of a UI tailored to the prospective use case.
  • Use of feedback and user testing to quickly iterate and turn a proof of concept into an MVP (minimal viable product).

Diagram: proposed AWS architecture of Daemon-designed knowledge support tool for Tachmed

The Outcome

• Empowering Innovation: Tachmed’s team now has AI-driven tools to discover clinical use cases and antibody candidates from scientific literature, aligning with their mission to deliver transformative healthcare solutions.

• Accelerated Development: The tool saves scientists hours on individual research tasks and up to a week per clinical investigation project per scientist, enabling faster iteration and innovation, reducing time-to-market for groundbreaking healthcare products.

• Real-World Use: Originally intended as a POC, the tool is already in active use, proving its immediate value and scalability.

• Foundation for Scale: With a modular, AI-driven architecture driven by the best-in-class Anthropic Claude foundation large language model, the system is poised for expansion, including automating vendor data scraping and scaling to larger datasets.

Testimonials

“Daemon’s expertise and collaboration have been instrumental in helping us build tools that align with our mission to transform healthcare data and diagnostics. The AI-powered system they developed is already saving us time, delivering valuable insights, and enabling us to innovate faster. It is the first step of using AI to revolutionise assay development and moving downstream to frontline remote, predictive and automated user applications. This partnership is a game-changer.” - Paul Christie (CEO)

Labels

• Healthcare

• Life Sciences

• Research Automation

• Antibody Discovery

• Generative AI

• AI / Artificial Intelligence

• Anthropic Claude Sonnet

• Amazon Bedrock

• AWS AppRunner

• Amazon Elasticache

• Amazon RDS