Welcome to The Blueprint, a new feature where we highlight how Google Cloud customers are tackling unique and common challenges across industries using the latest AI and cloud technologies. We hope to inspire others looking to innovate in their work.
The challenge:
Giles AI is a London-based startup that helps healthcare and life sciences organizations quickly extract insights from fragmented data, whether that data is available in an online repository (e.g. PubMed, NICE, the FDA etc.), local documents, or internal IP. Users can connect to internal and external data repositories and upload documents and images to the Giles AI platform; this integration allows users to the combined knowledge base for insights more quickly and efficiently, using natural language prompts and an intuitive interface.
As Giles AI grew in popularity, our incumbent cloud provider struggled to cope with complex data flows, new LLMs, and external APIs. Latency increased, slowing the user interface and impacting critical activities. Engineers also required a more agile development environment. Security is also a foundational feature of Giles AI and everything we build — with clinical, medical, and healthcare standards in mind, sensitive data must be protected at every step, both at rest and in transit.
The solution:
Giles AI leveraged Google Cloud’s modular, API-friendly, microservices-based architecture to minimize latency, easily manage complex clinical data flows in real-time, and capitalize on the latest and greatest AI foundation models as they are released.
Backend service orchestration in Google Kubernetes Engine and lightweight microservices in Cloud Run are complemented by specialized workloads in Compute Engine to keep the Giles AI platform available, flexible, and scalable without the heavy management and maintenance demands of legacy infrastructure. Cloud Load Balancing ensures efficiency.
Cloud SQL, Cloud Storage, and Document AI help the Giles AI platform manage structured and unstructured data and extract insights. Under the hood, Vertex AI handles model selection and prompt orchestration. The system is model-agnostic by design, enabling Giles AI to route queries to the most appropriate language model including hundreds available through Model Garden on Vertex AI.
With this highly flexible approach, Giles AI is able to deliver numerous healthcare and life sciences use cases from systematic literature reviews and regulatory reviews to meta-analyses, data extraction, and patient eligibility screening — all with high levels of accuracy and agreement.
To enhance security, we’re leveraging Cloud Armor to defend against Web-based attacks and Security Command Center to keep a close eye on its posture. Google Cloud regional databases help Giles AI localize data at rest — a critical need given healthcare regulations.
The architecture:
Source Credit: https://cloud.google.com/blog/products/ai-machine-learning/the-blueprint-giles-ai-transforming-medical-research-conversational-generative-ai/
