Production AI Infrastructure
A Meta-approved SaaS platform handling real client traffic since April 2026. Not a tutorial project — live in production.
See it in productionOpen to new roles
Hi, I'mIan Altemus
Cloud Infrastructure & DevOps Engineer
Building and operating production AI systems on AWS, Kubernetes, and Terraform with a security-first, least-privilege approach at every layer.
<60sAI response
5minSecure CI/CD pipelines with GHA
Event-driven on AWS
Zero-downtime deploys
IaC with Terraform
Autonomous AI handling live traffic
Prometheus & Grafana observability
Kubernetes with ArgoCD
Zero trust by design
<60sAI response
5minSecure CI/CD pipelines with GHA
Event-driven on AWS
Zero-downtime deploys
IaC with Terraform
Autonomous AI handling live traffic
Prometheus & Grafana observability
Kubernetes with ArgoCD
Zero trust by design
01What I do
What differentiates me
I design, deploy, and operate production cloud infrastructure with a security-first, least-privilege approach at every layer.
End-to-End Ownership
Infrastructure, networking, application layer, DevOps pipeline, and security scanning. From the first line of Terraform to production traffic.
See it in productionOperator Mindset
15 years running multi-location businesses before engineering. I understand what downtime actually costs the bottom line.
See it in production02Impact in production
<60s
Average DM response time. Down from 6 hours on the same workload.
27%
DM-to-booking conversion rate on live production traffic.
5min
CI/CD build time. Reduced from 20 minutes with caching and pipeline optimization.
106
Terraform-managed AWS resources deployed during a live migration with zero downtime.
03Portfolio
Let's have a look at my production work
Real systems handling real traffic. Every diagram below maps to infrastructure running today.
Featured production project
SilverLink AI
Live in productionAI PlatformMeta-approved multi-tenant AI SaaS automating Instagram DM responses for beauty and wellness businesses. Built the complete stack from scratch — infrastructure, networking, application layer, and DevOps pipeline. The AI receptionist responds in under 60 seconds, handles booking intent, and hands off to an agentic Playwright booking agent for live availability lookups. First client: Secretive Nail Bar, three Southern California locations.
ECS FargateAPI GatewaySQSDynamoDBTerraformGitHub ActionsClaude APIPlaywright
AWS architecture — live message flow
Instagram
Meta Graph API
API Gateway
HMAC verified
SQS
Queue + DLQ
ECS Fargate
Flask consumer
Claude API
Per-tenant voice
DynamoDB
Conversation memory
Response
Under 60 seconds
Key decisions: Provider abstraction layer bridges Claude API and AWS Bedrock for runtime switching without code changes. Agentic Playwright agent runs on a separate ECS task triggered by booking intent. Split CI/CD pipelines with path-based triggers so app changes never touch infrastructure pipelines.
Why SQS over direct processing?
- Decouples webhook ingestion from AI processing
- Prevents Meta webhook timeouts on slow AI responses
- Enables retry logic and dead letter queue strategy
Why ECS over EKS?
- Smaller operational footprint for a single-product platform
- Faster delivery without cluster management overhead
- Lower infrastructure cost at current traffic volume
Why DynamoDB over Postgres?
- Conversation access pattern fits key-value model exactly
- Zero operational burden, no connection pooling
- Native AWS integration with Terraform and IAM
Platform engineering project
Trendlink
Active buildPlatform EngineeringKubernetes-based prospect enrichment and intelligence platform spanning AWS and Azure. A Playwright agent aggregates and enriches real business data via Google Places API, scored and surfaced through a dashboard. Built on a hub-and-spoke multi-cloud topology with GitOps via ArgoCD, canary deployments via Argo Rollouts, and Vault sidecar secret injection.
KubernetesArgoCDArgo RolloutsHelmVaultPlaywrightAWSAzure
CI/CD pipeline — live deployment sequence
Git Push
main branch
GitHub Actions
build + scan
Helm Chart
packaged manifest
ArgoCD
detects drift
Argo Rollouts
canary split
Health Check
readiness probe
Promote
or auto-rollback
Pipeline logic: Every git push triggers GitHub Actions for build and security scanning. Helm packages the manifests. ArgoCD detects the new image tag and syncs. Argo Rollouts splits traffic at 20% canary. Health checks gate promotion. Failed probes trigger automatic rollback with no manual intervention.
Infrastructure topology
AWS/Azure
primary cloud
K8s Cluster
workload plane
Vault
sidecar injection
Playwright Agent
scrape + enrich
Google Places API
business data
Dashboard
scored results
Observability
Prometheus + Grafana
Pod and instance level metrics. Custom dashboards per workload.
Multi-cloud
AWS + Azure hub-and-spoke
Platform agnostic by design. Azure ACA planned for next phase.
Secrets
Vault sidecar injection
Secrets never in manifests or Git. Rotated independently of deployments.
Design principle: Infrastructure and application layers are fully decoupled. Terraform manages the platform. ArgoCD manages the workloads. Vault manages secrets. Each layer is independently replaceable without touching the others.
04Highlights
01
Built and operate a Meta-approved production AI SaaS platform on AWS, live client traffic, real business outcomes, zero production outages since launch.
02
Reduced average DM response time from 6 hours to under 60 seconds via event-driven architecture: API Gateway, SQS decoupling, ECS Fargate consumer.
03
Cut CI/CD pipeline build time from 20 minutes to 5 minutes using Docker layer caching, path-based triggers, and split app and infra pipelines.
04
Implemented GitOps with ArgoCD and Argo Rollouts, canary deployments, automated rollback on health check failure, Git as the single source of truth.
05
Hardened CI/CD security pipeline with Gitleaks secrets scanning, Bandit SAST, Trivy image scanning, and pip-audit required before every merge.
06
Designed and deployed an agentic AI booking integration using Playwright on managed ECS compute, triggered by booking intent detection in live conversation flow.
07
Systematic production debugging approach: establish blast radius first, isolate to network, application, or infrastructure layer, trace through logs and exit codes, remediate, then document root cause and update runbooks to prevent recurrence.
08
Designed and deployed a full observability stack using Prometheus sidecar injection and Grafana dashboards with custom pod and instance level metrics, moving beyond default cluster metrics to measure what actually matters for the application workload.
04Technical stack
Cloud
AWS
Cloud
Azure
Compute
ECS Fargate
Orchestration
Kubernetes
IaC
Terraform
GitOps
ArgoCD
CI/CD
GitHub Actions
Packaging
Helm
Secrets
Vault
Secrets
Secrets Manager
Messaging
SQS
Database
DynamoDB
Gateway
API Gateway
Observability
Prometheus
Dashboards
Grafana
AI
Claude API
Language
Python
Security
Trivy
Security
Bandit
Security
Gitleaks
Containers
Docker
Rollouts
Argo Rollouts
Automation
Playwright
Registry
ECR
05About
I am a Cloud Infrastructure and DevOps Engineer focused on AWS, Kubernetes, Terraform, and production AI systems. I design and operate infrastructure end-to-end with a security-first approach at every layer.
Before engineering, I spent 15 years running multi-location businesses. That background changed how I think about infrastructure. I know what it feels like from the stakeholder side when systems fail to meet objectives. I am not task-oriented. I am outcome-oriented at my core.
I am comfortable under pressure, meticulous about security, and genuinely enjoy debugging complex flows when things break. I combine operational discipline with clear, direct communication across technical and non-technical teams.
I have always been drawn to solving complex problems. Cloud and DevOps gave me the perfect outlet to realize that characteristic.
Certifications
- AWS Solutions Architect AssociateAmazon Web Services
- HashiCorp Terraform AssociateHashiCorp
- B.S. Cloud Computingin progressWGU
- CKA — Certified Kubernetes Administratorin progressCNCF
Currently open to
Cloud EngineerDevOps EngineerPlatform EngineerSRE
RemoteHybridOn-siteContract or FT
06Contact
Let's build reliable infrastructure together.
Open to Cloud Engineer, DevOps Engineer, Platform Engineer, and SRE roles. Remote, hybrid, or on-site. Costa Mesa, CA. Contract, contract-to-hire, or full time.