Upcoming Event: The AI Agent Implementation Framework
Upcoming Event: The AI Agent Implementation Framework
GE Healthcare
Bengaluru, India
Healthcare / Medical Devices
Amazon Bedrock | Amazon S3 | Amazon RDS | Amazon DynamoDB | Bedrock Agents | Microsoft Teams Tab Application | Azure AD / Entra ID | LLM-as-a-Judge Evaluation Framework | CI/CD Pipelines
GE Healthcare is a global leader in medical technology and diagnostics, operating complex, enterprise-scale AI systems to support its workforce worldwide. Facing a strategic shift away from Azure, GE Healthcare needed to migrate its internal AI agent platform, the Genie, to AWS without disrupting operations or losing functionality. Avahi led the full migration from Azure OpenAI to Amazon Bedrock, rebuilt the agent orchestration layer using AWS-native services, and delivered a new AI-powered work management agent directly into the live platform, all before the migration was complete. The result was a fully AWS-hosted Genie platform with uninterrupted service, validated model parity, and expanded capability.
GE Healthcare is a world-leading provider of medical imaging, diagnostics, patient monitoring, and healthcare IT solutions. Operating across more than 100 countries, the company develops technologies that enable clinicians to make faster, more informed decisions. To support its large internal workforce, GE Healthcare has invested significantly in enterprise AI tooling, including Genie, an internal AI agent platform used to streamline work management and information access across the organization.
GE Healthcare had built Genie, its internal AI agent platform, on Microsoft Azure using Azure OpenAI as the model provider. As the organization’s cloud strategy evolved, the team needed to migrate Genie fully to AWS, moving the AI model layer, agent orchestration, backend services, and all integrations to an AWS-native architecture.
The challenge was not simply technical. Genie was a live, employee-facing system. Any degradation in model behavior or service interruption during migration would directly impact the productivity of the teams that relied on it daily. Ensuring that Amazon Bedrock models performed with the same quality and consistency as Azure OpenAI, across thousands of potential agent interactions, required a rigorous validation approach, not guesswork.
At the same time, GE Healthcare wanted to extend Genie’s capabilities, not just preserve them. There was an opportunity to build and launch new features during the migration window, but doing so in a live enterprise environment, under compressed timelines, required both technical precision and strong delivery discipline.
The migration to AWS also presented an opportunity to modernize the platform’s deployment architecture, establishing automated, repeatable release processes aligned with enterprise engineering standards from the outset.
Had the migration not been addressed, GE Healthcare would have remained locked into Azure, foregoing the expanded model capabilities and operational advantages of Amazon Bedrock and the broader AWS ecosystem. The platform would have continued to accumulate technical dependencies on a stack misaligned with the organization’s cloud direction, increasing migration complexity and cost over time.
GE Healthcare’s decision to move to AWS was driven by a strategic desire to consolidate onto a cloud platform that offered a more expansive and rapidly evolving AI services ecosystem. Amazon Bedrock, in particular, provided access to a curated suite of foundation models and a robust agent orchestration framework that aligned with the long-term direction of the Genie platform.
Beyond the model layer, AWS offered the native infrastructure (S3, RDS, DynamoDB) to support a scalable, enterprise-grade backend without the overhead of managing cross-cloud dependencies. Migrating to AWS meant GE Healthcare could deepen its investment in a single, integrated cloud environment while gaining access to Bedrock’s growing model portfolio.
Avahi was selected for this engagement because of its demonstrated expertise in enterprise AI migrations and its standing as a premier-tier AWS partner. GE Healthcare needed a partner capable of operating in a complex, ambiguous environment, one where the platform had to remain live, the migration had to be seamless, and new features had to be delivered in parallel. Avahi brought the technical depth and delivery discipline to meet all three requirements simultaneously.
Critically, Avahi’s team understood that model behavior parity was not a checkbox. It was the foundation of stakeholder confidence. Rather than relying on manual review or informal comparison, Avahi introduced a structured, automated validation methodology purpose-built for this type of migration. That approach, combined with Avahi’s ability to deliver meaningful product value before the migration was complete, distinguished the engagement from a standard infrastructure lift-and-shift.
Avahi began with a thorough assessment of the existing Azure-hosted Genie architecture, mapping every dependency, integration, and service to its AWS equivalent. This blueprint informed every subsequent phase of the migration and ensured that nothing was left to assumption.
The most technically critical step was migrating the AI model layer from Azure OpenAI to Amazon Bedrock. To validate that model behavior remained consistent across the transition, Avahi implemented an LLM-as-a-Judge evaluation framework, a custom methodology that automated the comparison of agent outputs between the two platforms. Rather than manually reviewing thousands of agent responses, the framework systematically scored output quality and flagged divergence, all within the live Genie environment. This approach eliminated the guesswork from model validation and gave GE Healthcare’s stakeholders objective, evidence-based confidence in the migration outcome.
With the model layer validated, Avahi rebuilt the agent orchestration layer using Amazon Bedrock Agents, porting all backend services, APIs, and integrations to AWS-native infrastructure, including Amazon S3, Amazon RDS, and Amazon DynamoDB.
In parallel with the infrastructure migration, Avahi designed and built the Heartbeat agent, a conversational work management assistant, and integrated it directly into the live Genie platform ahead of the full migration’s completion. Delivering Heartbeat before the migration was finished demonstrated the team’s ability to create real product value under enterprise constraints, not just execute a technical handoff.
Avahi established a fully automated CI/CD pipeline for the AWS-hosted Genie platform, incorporating code quality guardrails and integrated security scanning at each release stage. Building this foundation as part of the migration ensured that Genie launched on AWS with an auditable, repeatable path to production — setting the engineering team up for sustainable, confident delivery going forward.
To embed Genie within GE Healthcare’s existing collaboration workflows, Avahi built and deployed a Microsoft Teams Tab application that surfaces the full Genie web application as an iFrame directly inside Teams. Rather than rebuilding the Genie UI or creating a parallel chatbot interface, this approach gave employees access to the complete, AWS-hosted Genie platform without leaving their daily Teams environment. Authentication was handled via Azure AD / Entra ID, enabling seamless SSO so users landed in Genie without a separate login step.
The engagement concluded with formal UAT conducted with GE Healthcare stakeholders, including managed Go/No-Go decisions and structured post-delivery feedback cycles to address any outstanding issues from the Heartbeat rollout.
Avahi delivered a successful, full-feature migration of GE Healthcare’s Genie platform from Azure to AWS, with no loss of service continuity and validated model behavior parity across the transition. The LLM-as-a-Judge framework replaced what would have been an unscalable manual QA process, giving GE Healthcare’s leadership the evidence needed to proceed with confidence at each stage of the rollout.
By building and launching the Heartbeat agent before the migration was complete, Avahi demonstrated that enterprise AI migrations need not be purely technical exercises. They can be vehicles for meaningful product advancement. GE Healthcare ended the engagement with an expanded, AWS-native AI platform and a repeatable integration pattern for future agent development.
Let’s explore your high-impact AI opportunities together in a complimentary session
