2026 — PRESENT
Shopee · Product
DevOps Engineer
Multi-agent incident diagnosis, governed self-healing Kubernetes, and a read-only engineering knowledge agent.
PRODUCTION AI SYSTEMS / DEVOPS / SRE
Making AI useful in production— without surrendering trust.
I'm Ian Juan. I build evidence-grounded, governed, and validated AI systems—from multi-agent incident diagnosis and human-approved remediation to a cited engineering knowledge agent that knows when to abstain.
139AI-SRE safety & regression tests
$6.5→$2.5Daily AI cost at comparable nonprod event volume
0Hybrid false-confidence in controlled adversarial cases
1,678Engineering knowledge chunks
FLAGSHIP SYSTEMS
The model reasons. The control plane decides.
Two shipped systems and four design blueprints form one governed software-delivery lifecycle. Models interpret context; evidence, authority, and final decisions stay in testable controls.
UNDERSTANDRELAYOPS
BUILDTICKETFORGE
SIMULATEOPSTWIN
DELIVERRELEASEPILOT
OPERATEAI-SRE
ENABLEDAYONE
01 / INCIDENT INTELLIGENCESHIPPED CASE STUDY ↗
AI-SRE
A governed loop from alert to verified recovery
Multi-agent RCA, evidence verification, deterministic policy, human approval, bounded remediation, and post-checks.
SIGNALSANALYZEVERIFYGOVERNACTPROVE
23 chaos scenarios27–30s rollback drill recovery
02 / WORK PROXYSHIPPED CASE STUDY ↗
RELAYOPS
An engineering proxy for the repetitive parts of work
It finds documentation and code, investigates live state read-only, and organizes context without impersonating a person or making authority decisions.
DOCS + LIVEPREFETCHINVESTIGATECITEANSWER
657 source files84 automated tests
03 / PROGRESSIVE DELIVERYCONCEPT / IN DEVELOPMENT
RELEASEPILOT
Evidence-driven rollouts
Change risk, dependency context, live SLOs, and deterministic go / hold / abort gates.04 / TICKET-TO-MRCONCEPT / IN DEVELOPMENT
TICKETFORGE
Proof-carrying changes
Ticket intake, isolated implementation, tests, sandbox proof, and a human-approved Draft MR.05 / DEVELOPER ONBOARDINGCONCEPT / IN DEVELOPMENT
DAYONE
A verifiable path to the first MR
Environment diagnosis, architecture context, workflow practice, and continuous onboarding improvement.06 / PRODUCTION DIGITAL TWINCONCEPT / IN DEVELOPMENT
OPSTWIN
Rehearse the future before production
Ephemeral change rehearsal with synthetic traffic, failures, migrations, and honest fidelity gaps.PROFILE
From delivery automation to resilient platforms.
My work spans CI/CD, multi-cloud migrations, multi-region GitOps, service mesh governance, observability, and production AI systems. I focus on turning recurring operational pain into reusable, testable platform capabilities.
Core tools include Kubernetes, GCP, Docker, Terraform, ArgoCD, Istio, Prometheus, and OpenTelemetry.
EXPERIENCE
Production engineering at scale.
2024 — 2025
Shopee · Business Intelligence
DevOps Engineer
Multi-region GitOps for 80+ apps, service mesh governance, observability and GPU cost optimization.
2023 — 2024
Gram
DevOps Engineer
VM-to-Kubernetes and cross-cloud migrations, delivery automation, CI/CD and cloud cost optimization.
2020 — 2023
WebComm Technology
Software Engineer
CI/CD from scratch, automated web and mobile releases, Java backend services.
MORE WORK
From core systems to useful products.
01 / AI CAREER TOOL
MATCHCV
Turns a resume and target job description into a focused gap analysis, tailored content, and interview preparation.
VIEW LIVE ↗02 / BROWSER GAMEStarwake Arena
A playable real-time, two-lane strategy card battle set across a floating ocean arena.
PLAY NOW ↗03 / TRAVEL UTILITYLooAround
A public-toilet discovery map with practical filters for accessibility, baby-changing facilities, cost, and opening hours.
OPEN MAP ↗04 / COMMUNITY PLATFORMCCOO Learning
Public site, member experience, admin operations, APIs, media, and production quality gates.
VIEW LIVE ↗05 / FIELD NOTESAI Ops Playbook
A practical framework for evidence-grounded LLM triage, guardrails, approvals, and observability.
READ GUIDE ↗