Aug 14, 2026
The pool was leaving clues
How AI found latent experiments in my Home Assistant telemetry, and what that reveals about the hidden value of operational data.
Jason Vallery · Builder and product leader
I build things
to understand them.
I have spent thirty years taking systems apart, making them work, and explaining what I found. These days that means cloud data, AI infrastructure, and a greenhouse in my backyard.
30 yearssince my first published byline
12½ yearsat Microsoft, ending in Azure Blob
Named by OpenAIunder Microsoft Azure Support
Longmont rootedBoulder County native, resident since nine
In progress
What I am building now
My day job, my home lab, and a greenhouse project with my son. Each one gives me a different problem to solve.
01
Cloud data
Product work at VAST Data
Working on the infrastructure beneath demanding data and AI systems.
See current work →02Agent systems
A private platform with guardrails
Operating a pod-per-repository fleet of coding agents on Kubernetes.
Explore the system →03Verdify
AI-assisted greenhouse control
Building a bounded control loop with my adult son, James.
Visit Verdify ↗A defining chapter · 2017–2025
Infrastructure beneath frontier AI
From 2017 through 2025, I led Azure Blob product work, including the Azure-side storage relationship supporting OpenAI and Microsoft AI. OpenAI’s 2019 Dota 2 report names me under Microsoft Azure Support. At SNIA SDC 2024, Jegan Devaraju and I explained how Azure Blob supported large-scale model training.
My part was product and infrastructure work, alongside a much larger team.
Read the Azure and OpenAI chapter“I came home from my first meeting with OpenAI in 2017… mind-blown.”
Writing
Notes from problems I wanted to solve.
The collection currently includes 40 articles: 33 technical pieces and 7 essays. Some solved an immediate problem; others captured what I believed at the time. I have left both kinds here.
Jul 10, 2013
Email marketshare
I used aggregate domain counts from MineMyMail to estimate webmail account share in 2013. The result measured addresses seen by the tool, not active users.
Nov 1, 2012
Why do we use Pulse at Neudesic? The four pillars of enterprise social networking
I reduced the enterprise-social use cases I had seen to four practical groups: culture, knowledge discovery, innovation, and identity and recognition.
Keep exploring