The work
Production systems, delivered by embedded pods.
Three recent engagements. Clients are confidential by agreement; the scope, stack, and outcomes are real.
Case 01 / National multi-vendor luxury retailer
A marketplace platform in 90 days, after a failed 9-month build
90 daysAI engineering pod
90 days
AI is most valuable inside an experienced engineering team, not as a replacement for one. Human judgment ran architecture and governance; AI is what compressed a failed 9-month effort into 90 days.
The challenge
- Fragmented vendor integrations, each on a different catalog format
- Manual onboarding and error-prone order routing
- Little operational visibility across vendors and orders
- A previous, unaffiliated 9-month effort had failed to deliver
What we did
- Embedded AI engineering pod: product, backend, frontend, QA, DevOps
- AI-assisted requirements, API design, code generation, and test generation
- Every feature began with a written specification approved by the client
Outcome
- Standardized vendor onboarding with automated catalog synchronization
- One unified API for every vendor through a middleware layer
- Faster engineering cycles with QA built into the pipeline
Google Cloud PlatformAPI-first servicesReusable vendor connectorsInfrastructure as CodeCentralized monitoring
Case 02 / Healthcare technology platform (HIPAA-regulated)
A secure order-intake foundation for a regulated platform
6 monthsAI engineering pod
Phase 1 won
In a regulated environment the first win isn't a feature, it's trust. Disciplined human review let the client validate the model with contained risk.
The challenge
- Needed a secure, working foundation for demos, POCs, and onboarding
- Compliance was non-negotiable: protected health information in scope
- Wanted to test an outsourced delivery model before committing further
What we did
- Pod owned core HIPAA-compliant infrastructure and the order-intake UI
- SSO, MFA, role-based access control, and centralized audit logging
- 100% human code review and an extra compliance checkpoint for anything touching PHI
Outcome
- Client's own meeting notes called the Phase 0 demo highly successful
- Proved the pod model on a HIPAA-sensitive platform without loosening compliance
- Led directly to a Phase 1 engagement for data modeling and back-office automation
React + TypeScriptGolang orchestration layerAmazon Aurora PostgreSQLAWSAuth0
Case 03 / Enterprise cloud and DevOps platform provider
A unified AI control plane without pulling the core team off roadmap
OngoingAI engineering pod
0 core engineers diverted
A new AI product surface doesn't have to cost you your roadmap. An embedded pod under your governance can move it forward in parallel.
The challenge
- Clients wanted centralized governance across multiple LLM providers
- Internal engineering was fully committed to the core platform
- No room to build a new product surface from scratch
What we did
- Embedded pod: product, platform engineering, DevOps and SRE
- Same spec-first governance as the client's own engineering organization
- AI accelerated scaffolding, API design, and test generation
Outcome
- A working gateway prototype with unified cross-provider visibility into AI usage and cost
- Replaced manual cost reconciliation across teams
- Validated the concept without diverting the core platform team
Kubernetes microservicesMulti-provider LLM gatewayUsage metering and cost attributionCentralized observabilityRBAC and audit trails
Next step
Your operation could be the next case study.
Thirty minutes with José. You bring the operation, we bring an honest read on where AI pays and where it doesn't.