Software Engineer · Open Source · Cloud

MD. Yunus

SRE who builds observable, self-healing cloud platforms.

Framework contributions merged into Cloudflare and Docker, Oracle ACE Associate. I make platforms observable, incidents short, and deploys boring.

Open to DevOps/SRE roles & OSS collaboration

View Projects Read the docs GitHub

engineer.yaml

# production-ready, merged upstream
name: "MD. Yunus"
role: devops-sre
cloud: [oci, aws, gcp]
observability: [prometheus, grafana]
merged_upstream:
  - cloudflare/vinext # framework SSR fixes
  - docker/agent # overflow auto-recovery
  - withcoral/coral # Discord source spec
  - gitlab/dependency-scanning # yarn berry parser
credentials: oracle-ace-associate
status: open_to_work

11+
PRs merged into Cloudflare, Docker, GitLab & Coral: 281
automated tests shipped to production: 45%
MTTR reduction via AI incident agent: 12+
OCI reference architectures published

01 · The Work

Infrastructure as code, literally

How I approach platforms — reproducible infra, burn-rate alerting, and AI-assisted incident response.

infra/oke-cluster.tf

# Reproducible infra — no console clicks
resource "oci_containerengine_cluster" "oke" {
  name               = "prod-oke"
  kubernetes_version = "v1.31"
  vcn_id             = module.network.vcn_id
}

alerts/slo-burn.yaml

# Alert on burn rate, not on noise
alert: ErrorBudgetBurn
expr: |
  sum(rate(http_requests_total{code=~"5.."}[5m]))
  / sum(rate(http_requests_total[5m])) > 0.02
for: 10m

agent/root_cause.py

# Correlate logs, metrics, traces — fast
async def investigate(incident: Incident):
    signals = await mcp.gather(
        logs, metrics, traces
    )
    return model.rank_hypotheses(signals)[0]

02 · The Difference

Feature comparison

What you get beyond the job description — receipts included.

Capability	Typical setup	What I ship
Open source	Lists repos on a resume	11+ PRs merged into Cloudflare, Docker, GitLab & Coral
Testing	Writes some unit tests	281 automated tests; coverage 34% → 98%
Incident response	Follows the runbook	AI root-cause agent — 45% faster MTTD
Observability	Installs a dashboard	SLO burn-rate alerts + instrumented services
Infrastructure	Clicks through the console	Terraform + GitOps (ArgoCD/Flux), reviewed in PRs
Credentials	Self-reported skills	Oracle ACE Associate, 4x OCI certified

03 · The Stack

Runs with your stack

Tools I ship with daily — from container runtime to alert routing.

Docker
Kubernetes
Helm
Terraform
Ansible
Argo CD
Cloudflare
Prometheus
Grafana
GitHub Actions
Python
TypeScript

Plus Oracle Cloud Infrastructure — Architect Associate, DevOps Professional, and Developer Professional certified.Read the documentation →

02 — Open Source

Contributions that ship

Eleven merged contributions across Cloudflare, Docker, GitLab, and Coral.

I didn't learn open source from a course. I learned it by reading framework internals and shipping fixes that other developers now build on.

04 — Featured Projects

Selected work

Problem, approach, and measurable impact.

Read the full case studies with tested numbers →

Incident Root-Cause Investigator Pro

PythonCoral MCPFastAPIDockerGitHub Actions

The problem.During incidents, engineers burn critical minutes manually correlating logs, metrics, and traces across disconnected tools while users stay impacted.

The work.Architected an enterprise AI agent orchestrating multi-source telemetry via a Coral MCP query layer. ML models analyze logs, metrics, and traces to surface actionable insights during incident response.

45% faster MTTD (simulated)
281 automated tests
failure-injection coverage

View on GitHub →

Production Observability Platform

JavaSpring BootPrometheusGrafanaGitHub Actions

The problem.Services shipping without instrumentation mean failures get reported by users instead of alerts — the most expensive way to find out.

The work.Built a 4-stage CI/CD pipeline (build, test, security scan, deploy) and instrumented a Spring Boot service with Prometheus metrics and Grafana dashboards, with automated synthetic testing and alert routing.

4-stage CI/CD pipeline
real-time dashboards
multi-channel alerting

View on GitHub →

Cloudflare Vinext Contributions

TypeScriptReactNext.jsNode.js

The problem.SSR gaps in a production framework blocked CSS-in-JS users (Apollo, styled-components, Emotion), and duplicated request-handling code hid security issues.

The work.Shipped framework-level fixes across SSR context, routing, module resolution, and test coverage — merged into a framework used in production at the edge.

7 PRs merged
1,627 duplicate lines removed
6 CodeQL issues fixed
0% → 100% module coverage

Notable details

PR #151: Exported `ServerInsertedHTMLContext` to unblock CSS-in-JS SSR for Apollo Client, styled-components, and Emotion users
PR #193: Documented ESM resolution for validator packages
PR #197: Fixed Dynamic require of .mjs on Node 22
PR #249: Wired `RouterContext.Provider` in Pages Router fixing `useRouter()` returning null
PR #258: Extracted shared request-handling utilities, removing 1,627 duplicated lines and fixing 6 CodeQL security issues
PR #345: Added snapshot tests for entry template generators (0% → 100% coverage)
PR #620: Prevented `useActionState` from becoming undefined after `redirect()`

View on GitHub →

Docker Agent Contributions

GoprotobufgRPC

The problem.LLM context-window overflows crashed agent sessions mid-task, and table text copied from the TUI came out garbled.

The work.Implemented proactive token compaction at 90% capacity with auto-recovery across LLM providers, and a visual-to-rune mapping algorithm for accurate text selection.

14 LLM provider patterns handled
33 integration tests
zero-crash overflow recovery

Notable details

PR #1973: Implemented proactive token compaction at 90% capacity and auto-recovery for overflow errors across 14 LLM provider patterns
PR #2175: Corrected table-text copy mismatches by implementing visual-to-rune mapping algorithm

View on GitHub →

06 — Writing & Publications

Field notes

Thoughts, post-mortems, and architectural notes scaled for public consumption — published onyunus25jmi.medium.com.

OCI reference architecturesKubernetes at scaledistributed systemsopen-source fixesobservability strategiessupply chain security

Apr 2026

Building a Netflix-Style IaC Platform on OCI: Resource Manager Stacks as Product APIs

Md Yunus

MD. Yunus

Infrastructure as code, literally

Feature comparison

Runs with your stack

Contributions that ship

Selected work

Incident Root-Cause Investigator Pro

Production Observability Platform

Cloudflare Vinext Contributions

Docker Agent Contributions

Field notes

Building a Netflix-Style IaC Platform on OCI: Resource Manager Stacks as Product APIs

Zero-Downtime Migration to Oracle Autonomous Database

How I Fixed Vite’s “Dynamic require of .mjs is not supported” Error in Cloudflare’s vinext

How I Fixed Apollo Client SSR in Cloudflare’s vinext by Wiring a Missing React Context

Building a Lightweight, Scalable Database Stack with Oracle Container Engine

How I Fixed ESM Resolution in Cloudflare’s vinext

When to NOT Use Kafka on OCI: A Cost and Complexity Analysis for Event Streaming Architecture

The Cold Start Problem Nobody Talks About (It’s Not Just Serverless)

Beyond NetworkPolicies: Infrastructure-Native Microsegmentation with Oracle Kubernetes Engine

Designing a 1 Million RPS Kubernetes Ingress Layer: From TCP Handshake to Kernel Tuning

Control Plane High Availability on OKE: What Oracle Actually Manages (And Why It Matters)

Scaling The Software Architecture: Beyond Linear Growth to Phase Transitions

Multi-Tenancy in Kubernetes: Soft vs. Hard Isolation

Observability Economics: Cost vs. Fidelity in Distributed Tracing

Securing the Software Supply Chain: Beyond CVE Scanning to Cryptographic Verification

The Split-Brain Dilemma: Architecting Consistency in Distributed Cloud Systems

The Day the Internet Stopped: Inside the AWS Outage of October 20, 2025

The Micro-Startup Stack: Engineering Velocity Without the Overhead

System Design Deep Dive: How to Build a URL Shortener That Won’t Crash at Scale

Write the next chapter