I'm a freelance Platform, Cloud, DevOps & Kubernetes Consultant building secure, automated cloud infrastructure, internal platforms, delivery systems, and AI/ML platforms.
This profile is a catalog of independent reference implementations inspired by real requirements I have encountered through consulting. To protect client confidentiality and intellectual property, the public projects are rebuilt with independently created infrastructure and public or synthetic dataβthey contain no client code, data, or resources.
Portfolio Β· LinkedIn Β· GitHub
- β Certified Kubernetes Security Specialist (CKS)
- β Certified Kubernetes Administrator (CKA)
- β AWS Certified Solutions Architect β Associate (SAA)
- β Red Hat Certified System Administrator (RHCSA)
| Capability | Typical challenges |
|---|---|
| Platform Engineering | Standardizing fragmented technology stacks, enabling self-service, enforcing guardrails, and supporting multiple teams on shared infrastructure |
| Cloud & Kubernetes | Designing AWS/Azure architectures, adopting EKS/AKS, automating infrastructure, and modernizing workloads |
| Delivery & Reliability | Implementing CI/CD and GitOps, controlling release risk, improving observability, and automating incident diagnosis |
| AI/ML Platforms | Building portable ML pipelines, provisioning GPU infrastructure, tracking experiments, and deploying models reliably |
Challenge: AI accelerates application delivery, but a single application flaw combined with misconfiguration can escalate into a breach spanning container, cluster, and cloud resources. How can teams keep delivery speed while containing breaches across every layer?
Implementation: Traced a React2Shell (CVE-2025-55182) exploit in a Next.js application on EKS across four layers β code, container, cluster, and cloud β from four perspectives: attacker, platform administrator, security specialist, and DevSecOps engineer. Covered exploitation, incident investigation and containment, layered hardening, and shift-left security in delivery.
Evidence: The attack chain reaches container secrets, Kubernetes resources, an internal API, and a cloud secret through an overly permissive IRSA role; replaying it after each hardening step β least-privilege IAM, restricted RBAC, NetworkPolicies, pod security, a patched non-root distroless image β shows exactly where access is blocked.
AWS EKS IRSA Kubernetes RBAC NetworkPolicy Pod Security CloudTrail GuardDuty Trivy Argo CD Terraform GitHub Actions
Live case study Β· Source code
Challenge: Help a growing organization standardize independently managed applications on a shared platform while preserving team autonomy, isolation, and governance.
Implementation: Built a GitOps-driven EKS platform with automated tenant onboarding, workload-specific compute, policy enforcement, secrets integration, TLS/DNS automation, and service-mesh networking.
Evidence: A tenant is defined through one configuration file; GitOps creates its platform boundaries and deploys an application to a live endpoint in approximately three minutes.
AWS EKS Terraform Argo CD Karpenter Istio External Secrets
Challenge: Reduce production deployment risk without sacrificing delivery speed or requiring operators to coordinate every release manually.
Implementation: Applied controls across three lifecycle phases: pre-release validation, progressive delivery, and post-release monitoring with automated operational notifications.
Evidence: The implementation demonstrates early validation, canary and blue-green delivery patterns, automated analysis, rollback controls, and fast release feedback.
Kubernetes Argo CD Argo Rollouts Grafana Slack
Live case study Β· Source code
Challenge: Close the gap between ML experimentation and reliable production use by connecting model development, platform infrastructure, and delivery automation.
Implementation: Built an end-to-end platform for a substituted computer-vision workload, integrating Kubeflow, MLflow, Katib, DVC, KServe, GPU infrastructure, IaC, and GitOps.
Evidence: The platform demonstrates the complete lifecycle from experimentation and training through model tracking, validation, deployment, and inference.
Kubeflow KServe MLflow Katib DVC AWS EKS Karpenter Terraform Argo CD
Live case study Β· Source code
Challenge: Reduce manual correlation of Kubernetes alerts, pod status, logs, events, and runbooks without giving an AI system unsafe production access.
Implementation: Built an event-driven diagnostic pipeline in which an AI agent receives alerts, investigates the cluster through read-only RBAC permissions, and sends findings to operators for review.
Evidence: The design separates observation, diagnosis, and remediation authority, keeping human operators in control of production changes.
Amazon Bedrock AgentCore AWS EKS CloudWatch EventBridge Lambda Terraform GitHub Actions
Live case study Β· Source code
The projects below explore common engineering questions that arise during platform, cloud, DevOps, and AI/ML engagements.
| Project | Engineering question | Stack |
|---|---|---|
| Kubernetes Deployment Playbook | How should teams choose among rolling, recreate, canary, blue-green, A/B, and shadow deployments? | AKS, Argo CD, Argo Rollouts, Istio, Kiali |
| Multi-Cloud Kubernetes | How can one application run across AWS and Azure with unified delivery, traffic control, and monitoring? | EKS, AKS, Terraform, Argo CD, Helm, Cloudflare |
| Istio Sidecar Mode | How do ingress, TLS, traffic management, canary delivery, A/B testing, and shadowing work with sidecars? | Istio, AKS, cert-manager |
| Istio Ambient Mode | How do the same service-mesh capabilities work with ztunnel and waypoint proxies instead of sidecars? | Istio, AKS, cert-manager |
| EKS Argo CD Capability | How can the EKS-managed Argo CD capability support cluster delivery? | EKS, Argo CD |
| AKS Argo CD with Helm | How can Argo CD be installed and managed consistently on AKS? | AKS, Argo CD, Helm |
| KServe GPU Inference | How can an LLM inference service run on GPU-backed Kubernetes infrastructure? | EKS, KServe, GPU nodes, Argo CD, Open WebUI |
| IP Exhaustion Mitigation | How can pod IP exhaustion in Amazon EKS be mitigated without causing downtime to existing workloads? | EKS, VPC |
| Project | Engineering question | Stack |
|---|---|---|
| Automated Architecture Benchmark β EKS | How much does each architecture decision improve behavior under identical load on EKS? | EKS, Terraform, GitHub Actions |
| Automated Architecture Benchmark β ECS | How much does each architecture decision improve behavior under identical load on ECS? | ECS, Terraform, GitHub Actions |
| Hybrid Cloud Data Warehouse | How can an on-premises data warehouse integrate with serverless AWS services? | PostgreSQL, AWS Lambda, API Gateway, S3 |
| VM-Based GitOps | How can GitOps practices extend to virtual-machine environments? | Ansible, Jenkins, EC2 |
| Multi-Environment Jenkins Pipeline | How can one pipeline provision and deploy multiple AWS environments consistently? | Jenkins, Terraform, AWS |
| Docker Multi-Stage Build | How much can multi-stage builds reduce Spring Boot and FastAPI container image sizes? | Docker |
| Project | Engineering question | Stack |
|---|---|---|
| SageMaker YOLO Pipeline | How can training, experiment tracking, quality gating, and serverless inference be automated on AWS? | SageMaker, MLflow, Lambda, CloudFront, S3, Terraform |
| Toronto Shared Bike MLOps | How can a forecasting model move through an automated training and deployment workflow? | SageMaker, Terraform |
| KubeTriage | How can an AI agent summarize Kubernetes logs and likely root causes while retaining human oversight? | Argo CD, n8n, Kagent, Alertmanager |
| Ollama CPU vs GPU | How does CPU and GPU inference performance differ across retrieval, generation, and memory tasks? | Ollama, NVIDIA GPU, EC2, Terraform |
| Project | Engineering question | Stack |
|---|---|---|
| Jenkins CI Pipeline with ECR | How can an in-cluster Jenkins pipeline build and publish container images to ECR? | Jenkins, Docker, ECR |
| Multi-Environment Jenkins Pipeline | How can infrastructure changes move consistently across multiple environments? | Jenkins, Terraform, AWS |
| Argo CD Notifications | How can inline and manifest-based notifications trigger Slack and GitHub Actions? | Argo CD, Slack, GitHub Actions |
Cloud: AWS, Microsoft Azure, Cloudflare
Platforms: Kubernetes, Amazon EKS, Azure AKS, Kubeflow, Karpenter, Istio
Infrastructure: Terraform, Helm, Kustomize, Ansible, Docker
Delivery: Argo CD, Argo Rollouts, GitHub Actions, Jenkins
Observability: Prometheus, Grafana, Alertmanager
Development: Python, Bash, Git, FastAPI
I can help with platform architecture, Kubernetes adoption, GitOps, cloud infrastructure automation, release engineering, and AI/ML platform development.
I'm available for freelance and consulting engagements.