whoami

Jay Honnold

Senior software engineer building platforms for AI agents. At home I run a homelab and local LLMs.

Senior Software Engineer @ Cognite // Phoenix, AZ · 33.45°N 112.07°W

cat about.md

Agents by day, homelab by night.

I'm a senior software engineer at Cognite, where I work on the platform behind Atlas AI: the agentic loop, model integrations, and the tools the agents call. Before that I spent five years at Amazon and AWS, mostly turning manual processes into automated systems. At AWS that work saved more than 50 developer-years.

At home I run a three-node Proxmox cluster the way I'd run production: configs in git with CI, monitoring, and encrypted offsite backups. It also hosts my local LLM stack, where I work on model serving, tracing, and agent memory on an RTX 4090. In my spare time I write Berserk, an open-source chess engine in C that ranks among the strongest in the world.

jay.tsTypeScript

export const jay = {
  role: "Senior SWE",
  focus: ["AI agents", "LLMs"],
  basedIn: "Phoenix, AZ",
  offHours: "homelab",
};

export const jay = {
  role: "Senior Software Engineer",
  focus: ["AI agents", "LLM platforms", "system design"],
  languages: ["TypeScript", "Java", "Kotlin", "Go", "Python", "C", "C++"],
  basedIn: "Phoenix, AZ",  // saguaro country
  offHours: ["homelab", "local LLMs", "Berserk"],
};

Jay Honnold standing on a snowy overlook

~/photos/overlook.jpg

github linkedin

git log --career --graph

Experience

5 commits · 2017 → now

9e1c0af
9e1c0af Mar 2026 — present
Senior Software Engineer · Cognite Cognite · HEAD → main HEAD → main
Working on the harness and platform behind Atlas AI, Cognite's low-code workbench for industrial AI agents. That covers the agentic loop, model integrations across Azure, AWS, and Google Cloud, Cognite Data Fusion (CDF) tools and the CDF MCP server, and services that back agent tools, such as a sandbox. Plus a little frontend work.
- LLMs
- MCP
- Azure
- AWS
- Google Cloud
Mar 2026 — present
a1f3c9e
a1f3c9e Apr 2022 — Mar 2026
SDE I → II → Senior Software Engineer · AWS AWS
Led internal automation for AWS global and regional expansion: validation services, code templates for expansion readiness, and scripts that use AI to refactor code for new regions.
- +1,300 pipelines supported
- +50 developer-years saved
- Java
- Kotlin
- TypeScript
- Ruby
- Go
- Python
Apr 2022 — Mar 2026
7be20d4
7be20d4 Feb 2021 — Apr 2022
SDE I · Amazon Amazon
Built and maintained order ingestion for Multi-Channel Fulfillment, handling thousands of orders a day. Also modernized legacy components and worked on the initial move to native AWS.
- Java
- Kotlin
- AWS
Feb 2021 — Apr 2022
3c91e02
3c91e02 Mar 2019 — Jan 2021
Full-Stack Developer · Allstate Allstate
Worked on a risk-assessment platform that checks 100+ components a day for production readiness. Led the split of its monolithic API into six microservices and rebuilt the UI to meet accessibility standards.
- +100 components assessed / day
- 1 monolith → 6 services
- React
- Java
- Spring
- Node.js
- MSSQL
Mar 2019 — Jan 2021
0d4a7b1
0d4a7b1 Jun 2017 — Aug 2018
Full-Stack Developer · Coder Inc. Coder Inc.
Built Victor, an Android app for veterans, and the Coder Platform, a web app that connects non-technical clients with developers.
- React Native
- React
- Rails
- PostgreSQL
Jun 2017 — Aug 2018

ls -la ~/works

Selected work

all repositories ↗

drwxr-xr-x homelab/ featured

$ cat ~/homelab/nodes
node     cpu     ram   gpu
alpha    3950X   32G   -
bravo    5950X  128G   4090
charlie  8500G   30G   -
$ ls ~/homelab/docs
hosts/  incidents/  plans/
$ ls ~/homelab/.gitea/workflows
lint.yaml  smoke.yaml

Homelab

Three-node Proxmox cluster run with the same practices I use on production systems at work.

+ every config lives in git, goes through a PR and CI, then deploys from the repo
+ Prometheus, Grafana, and Loki for metrics and logs, plus uptime checks
+ encrypted on-site and offsite backups, and written incident reviews
+ the always-on autonomous agent runs under its own least-privilege account

Proxmox
Docker
Prometheus
Grafana
CI/CD

private

drwxr-xr-x local-llm/

$ ls ~/models
qwen3.8-27B   gemma4-31B
gpt-oss-120b  deepseek-v4-flash
bge-m3
$ ls ~/agents
hermes  claude-code  opencode

Local LLM stack

Self-hosted model serving and agent infrastructure on an RTX 4090 host with 128 GB of RAM.

+ one OpenAI-compatible endpoint for many models, so tools switch models without code changes
+ LLM calls traced in Langfuse, with agent memory served over MCP