Corey Jones

Infrastructure Reliability Architect · Birmingham, Alabama

I build the reliability practice before you need one.

hello@coreyjones.dev · coreyjones.dev · github.com/unorig-username · linkedin.com/in/coreycjones

I'm Corey Jones. Twelve years in production systems, eight of them in SRE. I join early, stand up observability, Kubernetes, and incident response from zero, and make systems tell the truth about themselves. Right now I'm the sole SRE at Linq. I also build complete products, from Slack tooling to a self-hosted e-commerce platform.

Last checked just now Response time same day Uptime … On call since 2018

Page Corey Download résumé (PDF) GitHub

Metrics

At Linq · Jan 26, 2026 to Nov 15, 2026

1,243pull requests merged across 38 repos

415teammates' pull requests reviewed

22repos started, still the top contributor

8.4Mactive metric series on the stack I built

Merged pull requests per month Source: GitHub search

302

334

210

133

The August step is when I started running agent-driven development loops with review gates. November is partial, through the 15th.

Services

Last 60 days · hover a bar

Incident log

Newest first · all postmortems blameless

INC-2026-01Jan 2026 – nowmonitoring

Linq

Infrastructure Reliability Architect · sole SRE · reports to the CTO

Observability. Stood up the observability platform in my first two weeks: self-hosted Grafana, Mimir, Loki, Tempo, and Pyroscope on EKS, all in Terraform, fed by a two-tier OpenTelemetry Collector with tail sampling. It now carries about 8.4M active series, with roughly 140 alert rules and 560 recording rules in code. Moved the company's Grafana off Grafana Cloud and cut 55 datasources to 14.
Analytics platform. Built the ClickHouse platform end to end: signed HTTP ingest into NATS JetStream, then ClickHouse (three replicas on Graviton) and a Parquet archive in S3, on its own EKS cluster managed by Argo CD. Nightly backups, a staging mirror, and 10 runbooks. In production since September 2026.
Core platform. 332 merged pull requests in the core messaging platform (Go): Postgres vacuum and WAL tuning, about 123 GB of redundant indexes dropped, NATS JetStream capacity and placement work, send-path caching and fetch fixes, and the removal of the Kong API gateway with zero partner errors at cutover.
Device management. Designed and built a self-hosted Apple MDM (NanoMDM, zero-touch Apple Business enrollment, step-ca) plus its operator UI in the internal Tech Ops platform. Zero-touch enrollment proven in October; production pilot running in November.
Platform foundations. A self-hosted GitHub Actions runner fleet on EKS (ARC and Karpenter) that cut lint from 127s to 31s and queue time from 67s to 3s. An AWS Organizations landing zone in OpenTofu: account vending, SCP guardrails, Identity Center with SCIM, org-wide CloudTrail and GuardDuty, and GitHub OIDC with no static keys.
Incident response. Configured incident.io as code (severities, service catalog, alert routing, Linear and Notion workflows). Built the partner paging bot, a status-page-to-incident bridge, a heartbeat monitoring service, and Megaphone for broadcasts and partner alerts.
Prototypes. TraceLinq, a Rollbar replacement built in seven days (Go API, React dashboard, four SDKs, Helm chart). Siren, a Slack incident app with Claude-drafted status updates. Specter, a live map of edge traffic.

INC-2020-12Dec 2020 – Aug 2025resolved

Fetch

Czar of Site Reliability Engineering · first SRE hire

Joined as the first SRE and built the SRE team from there.
Built, ran, and migrated the company onto self-hosted Grafana clusters, Loki (replacing several log aggregators), and Mimir for Prometheus metrics, running at 22.5M active series.

INC-2018-01Jan 2018 – Dec 2020resolved

Shipt

QA Engineer → Site Reliability Engineer → Senior Site Reliability Engineer

Built and ran custom incident-management software integrated with Slack.
Moved observability from hosted Graphite to a self-hosted Graphite cluster on Kubernetes with a carbon-relay tier, then to self-hosted Prometheus. Ran Grafana in Kubernetes.
Architected the load-testing platform, built Jenkins CI for integration tests, and ran a Selenium grid in Kubernetes.
Kept retail partners' S3 and SFTP integrations reliable, and taught classes in Shipt Tech University.

INC-2014-11Nov 2014 – Jan 2018resolved

Influence Health (formerly MEDSEEK)

Systems Integration Engineer

Built integrations that normalized HL7, C-CDA, and FHIR messages.
Deployed and upgraded the patient-portal platform for hospital systems, with on-site work for sales, architecture, and support.

INC-0000-00Recurringpattern

Root cause: arrives early, builds it right

Contributing factors: information, logistics, accuracy

First SRE hire at Fetch. Slack incident tooling at Shipt in 2019 and again at Linq in 2026. Self-hosted Grafana stacks at three companies. I show up before the reliability practice exists and leave it running.

Things I've shipped

Built, not just configured

Linq repositories are private, so those cards describe the work without links. Numbers come from the repositories themselves.

Observability · Linqproduction

LGTM+P observability platform

Metrics, logs, traces, and profiles in one self-hosted stack, fed by OpenTelemetry and run entirely from Terraform and code-reviewed rules.

8.4Mactive series

~700alert and recording rules in code

2 wkto first light

OpenTelemetry · Grafana · Mimir · Loki · Tempo · Pyroscope · Alloy · EKS · Karpenter · KEDA · Terraform

Data platform · Linqproduction

ClickHouse analytics platform

Signed ingest, a JetStream buffer, ClickHouse for queries, and a Parquet archive for the long tail. Built, documented, and handed to on-call with runbooks.

3ClickHouse replicas on Graviton

395 dParquet archive retention

10runbooks

Go · NATS JetStream · ClickHouse (Altinity operator) · S3 Parquet · EKS · Argo CD · Terraform

Device management · Linqpilot

Self-hosted Apple MDM

Zero-touch enrollment and declarative configuration for company devices, with a device API and an operator UI that refuses writes unless it owns the device.

0hands needed to enroll a device

17 dfrom first enrollment to production pilot

1operator UI

NanoMDM · NanoDEP · step-ca · Python · RDS MySQL · Kustomize · OpenTofu · React

Slack tooling · Linqproduction

Megaphone

Company-wide broadcasts that can be edited everywhere after sending, partner alerts for flagged and recovered lines, status-page fan-out, and micro-incidents.

~50klines of code

50test files

83%of commits mine

Python · FastAPI · Celery · Postgres · Redis · Slack Bolt · React · Helm · Terraform

CI · Linqproduction

Self-hosted CI runner fleet

Org-wide GitHub Actions runners on EKS with arm64 and x64 pools, scale to zero, and warm minimums during the work week.

4×faster lint (127s → 31s)

3squeue time, down from 67s

200runner ceiling on the busiest tier

Actions Runner Controller · EKS · Karpenter · OIDC · Terraform

On-call · Linqproduction

Paging and on-call plumbing

incident.io as code, a Slack bot that lets partners page on-call, a bridge from status-page webhooks to incident alerts, and a heartbeat monitoring service.

4services built

2 dto build and ship the paging bot

95test files in the heartbeat service

incident.io · Terraform · Python · Go · FastAPI · Flask · Slack Bolt · Helm

Visualization · Linqrunning

Specter

A live world map of HTTP traffic: each request is an arc from client to edge location to origin, colored by outcome. It runs on the office wall displays.

1Go binary, page embedded

0databases

0client IPs stored

Go · JavaScript · Cloudflare Logpush and Instant Logs · Helm

Error tracking · Linqprototype

TraceLinq

A self-hosted Rollbar replacement: fingerprint grouping, releases and deploy markers, Slack, PagerDuty, Linear, and GitHub integrations, and Claude root-cause notes.

7days to build

4SDKs: Go, Node, Python, Swift

65user stories shipped

Go · React · Vite · Postgres · Helm · Terraform

Developer tooling · Linqbuilt

Xavier

A versioned source system for agent-driven development: it bootstraps repos with review lenses, deterministic gates, and human approval, and tracks loop health across every repo.

136merged pull requests

76test files

100%of human commits mine

Go · shell · GitHub Actions · Grafana

E-commerce platformlive

Naptime Candles

A custom storefront and admin portal for a handcrafted soy candle business, replacing Shopify end to end: Stripe checkout with server-side price checks, Shippo labels, gift cards, local delivery, and pickup.

34klines of app code

79test suites

$0Shopify subscription

Astro · React · Cloudflare Workers · D1 · R2 · Stripe · Shippo · Resend · Grafana

naptimecandles.com ↗

Performanceranked

Server challenges

Public load-test contests: implement one API, get benchmarked on a fixed box. I entered in C, Rust, Go, Zig, x86-64 assembly, Java, and, for fun, COBOL, Emacs Lisp, and Neovim Lua.

1Musers held on 16 cores

#5best leaderboard rank

32pull requests merged upstream

C · epoll · io_uring · Postgres wire protocol · SQLite · Rust · Zig · asm

See the results ↓

Computabilityresearch

Busy Beaver tooling

Go tooling for the open BB(6) holdouts: a machine parser, a simulator checked against the BB(5) champion, a run-length accelerated simulator, and cycler deciders.

47.2Msteps, BB(5) reproduced exactly

2decider families

48commits

Go · bbchallenge corpus

Local AIlab

Local-first AI lab

A private RAG knowledge base on Gemma 4, Whisper transcription with speaker labels on Apple Silicon, and a blind-graded comparison of shared agent-memory systems over MCP.

0tokens sent to the cloud

36eval questions, graded blind

3memory systems compared

Python · Ollama · LlamaIndex · Chroma · MLX Whisper · Mem0 · Graphiti · MCP

Load tests

Public leaderboards · as of Nov 15, 2026

Same box, same load generator, same rules for every entry. My best entry in each is C written from the socket up: an event loop per core, hand-written HTTP/1.1, and a hand-written Postgres wire client where there is a database.

Challenge	Box	Result	Rank
Million usersc-epoll · PR #104	16 vCPU, Postgres on its own box	1,000,000 users (max)	#8 of 85
Multicore serverc-epoll · PR #203	8 vCPU, Postgres on the same box	210,000 users · 52k req/s	#5 of 127
$12 serverc-iouring · PR #290	1 vCPU, 2 GB, SQLite	280,000 users · 27k req/s	#10 of 237

Novelty entries

Same $12 box, same load test, in languages nobody should run a web server in. Every one of them passed the test.

Entry	Result	Rank
x86-64 assemblyHTTP, JSON, SHA-256, and JWT in assembly; raw io_uring syscalls · PR #206	250,000 users	#33 of 237
COBOLGnuCOBOL; HTTP server, router, JSON, and JWT all in COBOL · PR #185	126,000 users	#101 of 237
Lua inside Neovimnvim --headless, 300 lines of Lua, libuv, SQLite over LuaJIT FFI · PR #443	110,000 users	#111 of 237
Emacs Lispemacs --batch parses HTTP and checks JWTs; built-in SQLite, behind Nginx · PR #444	24,000 users	#208 of 237

Dependencies

What I run in production

Observability

OpenTelemetry · Grafana · Mimir · Loki · Tempo · Pyroscope · Alloy · Prometheus · Graphite

Platform

Kubernetes (EKS) · Karpenter · KEDA · Helm · Argo CD · Terraform · OpenTofu · AWS · Cloudflare Workers

Data

PostgreSQL · ClickHouse · NATS JetStream · Valkey/Redis · SQLite and D1 · S3 and Parquet

Languages

Go · Python · TypeScript · C · Rust · Zig · SQL · shell

Incident response

incident.io · PagerDuty · Slack apps · Linear · status pages · blameless postmortems

AI engineering

Claude and Bedrock · MCP servers · agent loops with review gates · local models (Ollama, Gemma, Whisper)