Project Overview

LucidAct, a leader in remote patient monitoring, sought a way to automatically generate critical reports and detect anomalies in patient vital data over time. Avahi helped LucidAct build a proof of concept (POC) that
leverages AWS Generative AI services to automate report creation and streamline data analysis. This cloud-based solution enables care providers to quickly pinpoint important health changes, ensuring more proactive and data-driven patient care.

About the Customer

LucidAct operates in the healthcare technology sector, specializing in remote patient monitoring and digital health solutions. Their core offerings enable healthcare providers to track patients’ vital statistics in real-time, aiming to improve patient outcomes and reduce the burden on clinical resources.

The Problem

LucidAct needed to automate the generation of patient health summaries and anomaly detection. Manually compiling these reports consumed valuable clinical time, and the risk of overlooking subtle trends in large datasets was high. Without addressing this challenge, their customers (healthcare providers) would continue to rely on time-intensive manual processes, potentially slowing response times to critical changes in patient health.

Why AWS

AWS offered LucidAct a mature ecosystem of AI and machine learning services such as Amazon Bedrock, Amazon Sagemaker, and AWS Transcribe, which simplified data ingestion, model training, and natural language processing. The global reach and managed services provided by AWS ensured high availability, scalability, and security—key requirements for handling sensitive patient data at any scale.

Why LucidAct Chose Avahi

Avahi’s proven track record in designing AI-driven solutions on AWS made them a clear choice for LucidAct’s ambitious POC. As an AWS Premier Tier partner, Avahi brought expert guidance on choosing the right combination of AWS services for rapid development and seamless integration. Avahi’s team of senior AI architects and data scientists offered deep experience with Generative AI pipelines, ensuring a smooth and insightful collaboration.

Solution

Avahi led a four-phase approach to deliver a comprehensive end-to-end POC for LucidAct. First, a discovery and planning phase mapped existing MongoDB data flows and finalized success criteria. Next, Avahi built the core architecture using AWS Lambda for event-driven processing and Amazon S3 for storing patient data and generated reports. AWS Sagemaker and Amazon Bedrock were implemented to power the natural language summaries and detect anomalies in vital data over time.

Additionally, AWS Transcribe was introduced for any needed speech-to-text capabilities (such as voice notes from clinicians). An Amazon API Gateway fronted the solution, allowing external applications to securely request summaries. Lambda functions handled indexing, embedding creation, and updates to ensure LucidAct’s data was always ready for Generative AI inference. Throughout the project, CloudWatch metrics monitored pipeline performance, usage, and potential errors.

Once the pipeline was validated, Avahi delivered a comprehensive runbook to help LucidAct manage deployments and future updates. The final handoff included a detailed presentation, documentation, and knowledge-transfer sessions so LucidAct could maintain and extend the solution in-house.

• End-to-end POC for report generation and anomaly detection on patient vital data
• Automated embeddings/index creation for text and tabular data
• Comprehensive runbook for deployments and operational readiness
• Data processing pipeline with event-driven Lambda functions
• Integration with AWS Bedrock and Sagemaker for Generative AI models
• Final documentation, presentation, and knowledge transfer sessions

Project Impact

By automating patient reporting and integrating advanced anomaly detection, LucidAct gained a faster, more accurate system for providing critical health insights. Clinicians can now focus on proactive patient care rather than manual data aggregation.

• Validated accuracy, bias, and scalability of Generative AI models using standard tools
• Streamlined data ingestion and report generation for all patient vitals stored in MongoDB
• Established an automated pipeline that reduces manual reporting efforts significantly