Eric Chiu

Cloud / DevOps / SRE

I build AWS infrastructure as code — and the pipelines that ship it.

Recent Computer Engineering grad and AWS-certified architect, with 6 years leading frontline operations. My standout build is an AI-assisted SRE pipeline on Step Functions, Bedrock, and Lambda that triages incidents in about six seconds.

View résumé See the AI-SRE project

Featured Projects

Flagship

AI-Assisted SRE Incident Analysis

Step FunctionsBedrockLambda (ARM64)EventBridgeDynamoDBTerraformPython 3.11

Event-driven serverless pipeline that catches infrastructure alarms, fans out three parallel collectors (metrics, logs, deploy context) through Step Functions, and hands the evidence to Amazon Bedrock for root-cause hypotheses — advisory-only, with explicit IAM denies so the model can never take an action.

~6send-to-end triage
~$5/moper 100 incidents
90dincident store · TTL
  • Express Step Functions for sub-second, high-volume fan-out over Standard.
  • ARM64 Lambda for ~20% lower compute cost at equal throughput.
  • Advisory-only by design — IAM explicitly denies the analyzer any mutating action.

Enterprise 3-Tier App on ECS Fargate

ECS FargateRDS PostgreSQLLGTMTerraformDocker

Refactored a 1,200-line monolith into 9 focused services behind an ALB, with a full LGTM observability stack.

React + Node.js + PostgreSQL 15 (Multi-AZ) on ECS Fargate behind an ALB with full VPC isolation. Grafana/Loki/Tempo/Mimir observability, Secrets Manager, WAF, and SSL/TLS termination at the edge.

  • 3-tier isolation: ALB → Fargate frontend → Fargate backend → RDS in private subnets
  • Event-driven pool metrics instead of polling; health checks and alerting throughout
  • Cost-tuned: Fargate Spot, RDS auto-pause, log-retention policies

Aura — EKS GPU Autoscaler

EKSKarpenterKubernetesTerraformGitHub Actions

Ephemeral GPU clusters that scale to zero — no idle g5.xlarge burning ~$1/hr between jobs.

Modular Terraform + GitHub Actions system for on-demand EKS clusters: deploy, run a batch AI workload, collect results, then self-destruct. Karpenter provisions CPU and GPU NodePools per demand.

  • Full lifecycle in one pipeline: Deploy → Run → Collect → Destroy
  • Karpenter NodePools sized for LLM inference; consolidation on a 60s window
  • OIDC identity federation — GitHub Actions as trusted IdP, zero static AWS keys

FinOps Zombie Hunter

LambdaEventBridgeTerraformPython 3.12tfsec

Weekly sweep that flagged ~$230/mo of orphaned resources in a test account.

Scheduled cost-optimization engine that hunts orphaned "zombie" resources (EBS volumes, NAT Gateways, RDS, Elastic IPs) across every enabled region, on a Sunday EventBridge cron.

  • CloudWatch usage-based detection — flags by real usage, not just status
  • Dry-run by default; two-tier report-only vs. delete safety model
  • Shift-left security: tfsec + flake8 gate the pipeline before any deploy

Cloud Resume Challenge This site

S3CloudFrontLambdaDynamoDBTerraform

The page you're on. Static assets from S3 via CloudFront OAC; the visitor count below is served live by a Python Lambda + DynamoDB behind API Gateway — full IaC, OIDC-based CI/CD, custom domain.

— views · live from Lambda + DynamoDB

Plus MCP DevOps Mentor and more — all repositories on GitHub →

Stack

Cloud · AWS

  • ECS Fargate
  • EKS
  • Lambda
  • Step Functions
  • EventBridge
  • Bedrock
  • RDS
  • DynamoDB
  • S3
  • CloudFront
  • API Gateway
  • VPC
  • IAM
  • WAF
  • Secrets Manager

IaC & CI/CD

  • Terraform
  • GitHub Actions
  • OIDC federation
  • CloudFormation

Containers & Orchestration

  • Docker
  • Kubernetes
  • Karpenter

Observability

  • Grafana
  • Loki
  • Tempo
  • Mimir
  • Prometheus
  • CloudWatch
  • X-Ray

Security & FinOps

  • IAM least-privilege
  • tfsec
  • Cost automation
  • Right-sizing

Languages

  • Python
  • Bash
  • HCL
  • YAML
  • SQL
  • Linux

Before Cloud — Operations Leadership

Six years owning outcomes on the floor: real-time incident triage, escalation ownership, and team coordination under pressure. The on-call mindset, before I had the word for it.

  1. Shift Leader

    Teazzi 2021 – 2024
    • Ran live operations for a high-volume location — real-time triage of equipment failures, supply shortages, and customer escalations under rush-hour load.
    • Built the closing and shift-handoff checklist that cut transition errors and kept the next shift unblocked.
    • Directed a 4–6 person team through peak throughput, reassigning roles on the fly to hold service standards.

  2. Assistant Manager

    OneZo 2018 – 2021
    • On-site decision-maker in the manager's absence — triaged issues in real time and kept operations running.
    • Supervised 6–10 staff: scheduling, onboarding, and SOP training, building documentation and knowledge-transfer habits.
    • Partnered with ownership on inventory, staffing, and quality control against tight timelines.

Education & Certifications

Education

B.S. Computer Engineering

University of Massachusetts Boston

Certifications

  • AWS Certified Solutions Architect – Associate SAA-C03
  • AWS Certified Cloud Practitioner CLF-C02

Currently preparing: HashiCorp Terraform Associate

Open to Cloud / DevOps / SRE roles

Boston or remote. The fastest way to reach me is email — résumé is one click away.

ericchiu0228@gmail.com View résumé