Zyla Re-Platforms from AWS to GCP in Six Weeks with Facets

How Zyla leveraged Facets' Blueprints and Praxis AI agents to complete a full AWS-to-GCP production re-platform in 6 weeks with zero data loss and a zero-downtime DNS cutover.

About Zyla

Zyla operates a healthtech platform serving consumer-facing apps and partner integrations across the Indian healthcare ecosystem. Its production workload powers clinical workflows, patient records, doctor onboarding, wallets, and HRA services, backed by relational databases, terabytes of stored documents and images, real-time messaging, and a mobile experience delivered through a global content network.

The Challenge: Full AWS-to-GCP Re-Platform with a Small Team

Zyla made the strategic decision to take its production workload off AWS entirely and move to Google Cloud — a full re-platform, not a hybrid step. The business stakes:

• A fixed business timeline. The migration had a defined deadline tied to platform consolidation goals.
• No room for downtime or data loss. The platform serves doctors, patients, and partners on a daily basis.
• A small platform team. Zyla could not afford to pull engineers off product work for a multi-quarter migration.
• A complete stack to move. Every layer — compute, databases, storage, networking, security, CI/CD — needed a safe path across.

Manual lift-and-shift for a stack this wide is typically measured in quarters, not weeks. The risk was not technical feasibility; GCP equivalents exist for everything. The risk was speed, correctness, and execution discipline. Any one of the three slips would have meant a missed timeline, lost product velocity, or — worst of all — lost customer data.

The Solution: Blueprint-Driven Orchestration + Praxis AI Agents

Phase 1: Blueprint Design and Staging Environment

Facets started by modeling Zyla's target GCP infrastructure as code — a single declarative blueprint that became the source of truth for both staging and production:

• AI-Led Discovery. Praxis AI agents mapped Zyla's entire AWS source state in days, eliminating weeks of manual auditing.
• Staging Live in Week One. A fully working GCP environment was up within one week, giving Zyla's team something to validate against immediately.

Phase 2: Cross-Cloud Bridges

While staging stabilized, Facets engineered the bridges that would carry Zyla's data and assets across clouds without disruption:

• Live Database Replication. All databases continuously synchronized to GCP, keeping AWS and GCP in lockstep until cutover.
• Storage Synchronization with Cost Discipline. 5 TB of stored content moved to GCP. Facets coordinated the storage sync with an AWS Egress Fee Waiver secured for Zyla, eliminating migration egress costs.
• Identity, Access, and Integrations. External services and image registries mirrored across clouds, with Zyla's existing security model carried forward intact.

Phase 3: A Rehearsed, Reversible Production Cutover

Cutover followed a published Day 1 / Day 2 plan reviewed and approved with Zyla in advance — predictable, rehearsed, and reversible:

• Zero-Downtime DNS Flip. Every public-facing record validated against the new environment before the switch — customers saw no disruption at all.
• Database Cutover. Final data sync across all databases, with continuous integrity checks throughout.

AI Agents: The Migration Accelerator

Praxis is the AI agent ecosystem inside Facets. Our engineers worked with Praxis agents to compress what typically takes quarters of work into days:

• AWS source state is mapped in hours, not weeks, eliminating the slowest step of any migration.
• Target environment modeled as code automatically, not built by hand.
• Migration scripts, validation passes, and runbooks drafted by agents and reviewed by engineers — quality and pace at the same time.
• Live root-cause analysis after go-live. When an issue surfaced post-cutover, a data migration agent built over Praxis pinpointed it to a pre-existing application bug, not the migration itself — saving days of customer investigation.

The result: a small platform team operated with the effective throughput of a much larger one, for the duration of the engagement.

"Praxis caught things in hours that would have taken our team days to find. But the part I trusted was that an engineer had read it before anything was applied to our environment." - Dilip Singhal, DevOps Engineer, Zyla

"

Strategic Value Delivered by Facets