AI-Assisted Support Automation | CI/CD Reliability | Python

I build reliable support and operations workflows with Python, APIs, CI/CD, human approval, and evidence-driven escalation.

My background combines enterprise technical support, Linux infrastructure, incident ownership, and developer-platform troubleshooting. My public projects use synthetic data and standalone implementations so I can demonstrate product thinking and engineering practices without exposing employer or customer information.

A production-shaped SRE and developer-platform project for ephemeral GitHub Actions runner fleets:

• OpenTofu configuration for a temporary AWS EKS lab
• Kubernetes, Helm, and Actions Runner Controller configuration
• Queue, startup, cleanup, and maximum-age SLOs
• Prometheus metrics and alert rules
• Capacity, cost, and incident evidence
• Read-only MCP tools for fleet inspection and recommendations
• Deterministic burst, capacity-loss, image-failure, and API-degradation tests

Current status: validated local and infrastructure scaffold; cloud deployment evidence has not yet been captured.

Tested Python tools and synthetic case-study patterns for:

• Evidence-led incident triage and next diagnostic steps
• Engineering-ready escalation quality checks
• Clear separation of observations, hypotheses, and root cause
• Incident communication, postmortems, and operational documentation

A read-only Python CLI that converts Linux fleet inventory into repeatable operational checks and guarded server-decommission plans:

• Patching, backup, monitoring, encryption, and OS-support readiness checks
• Explicit validation, prioritized findings, and JSON output
• Safety gates for approvals, dependencies, recovery evidence, and production
• Synthetic infrastructure data, unit tests, and CI-friendly exit codes

Policy-gated, approval-bound, staged Ansible rollouts for Linux fleets, shown through an SSSD directory migration:

• Planner that runs a staging canary first and never splits HA pairs into one wave
• Policy gate for wave size, freeze windows, readiness, and staging-before-production
• Approvals bound to a plan digest, with no self-approval and in-order wave apply
• Ansible role with automatic rollback, tested by Molecule on Rocky Linux 9 and Ubuntu 24.04

A Python CLI for repeatable CI/CD security and reliability reviews:

• Least-privilege permissions and immutable action references
• Unsafe untrusted-input and privileged pull-request patterns
• Timeouts, concurrency, OIDC, and reusable-workflow boundaries
• Human-readable and JSON output for local checks and CI

A dependency-free Python CLI that checks whether a support escalation still has a clear path to a customer outcome:

• Technical and customer-impact ownership, update cadence, and OOO coverage
• Business-day engineering response targets by severity
• Inactivity closures that are mislabeled as customer recovery
• Reopened escalations without a fresh technical decision

A gh CLI extension for a fast, read-only first look at a GitHub Enterprise Server support bundle: disk, memory and OOM, failed services, and proxy or connectivity errors.

• Support Knowledge Hygiene: score knowledge assets for ownership, freshness, usage, and duplication risk.
• Code Scanning Merge-Protection Reproduction: a minimal reproduction of analysis selection across Git references.

• Make the problem testable. Start with immutable evidence, define the failure boundary, and distinguish facts from hypotheses.
• Keep people in control. Require explicit review before generated content changes a customer-facing workflow.
• Build for reuse. Convert recurring investigations into tested tools, small reproductions, and practical documentation.
• Measure reliability. Evaluate expected behavior, unsafe claims, latency, and cost rather than relying on a polished demonstration alone.
• Improve the handoff. Give engineering a focused question, the evidence needed to answer it, and a clear customer-impact statement.
• Communicate uncertainty honestly. A useful outcome can be a bounded next step, not a premature root-cause claim.

Python · FastAPI · OpenTofu · Kubernetes · Helm · Prometheus · MCP · SLOs · GitHub Actions · CI/CD · Linux · Docker · APIs · Incident Response

• Secure and Reliable GitHub Actions Workflows
• GitHub Enterprise Server Operational Readiness
• GitHub Advanced Security Rollout and Operations
• From Symptom to Engineering-Ready Escalation
• Closure Is Not Customer Recovery
• Synthetic Support Engineering Case Studies

These are personal, unofficial projects. Public examples use synthetic data and do not contain employer source code, customer information, support-case data, or internal documentation.