Joshua Puetz / Healthcare IT · Networking · AI applications

The problem isn’talways whereyou first see it.

I’m Joshua Puetz. In healthcare IT and networking, I kept seeing issues arrive as “network problems” when the cause was somewhere else. I started building tools to make it easier to see what was actually happening.

01 / Selected work

A closer look.

Follow an outage, compare wireless readings, or keep working on an assistant’s answer. These are the tools I’ve been building, with the decisions and checks behind each one.
Try the app demos

Infrastructure / Local AI

Wireless Observability

Local lab · verified recordings

I built a Python layer that joins Grafana alerts with read-only checks across storage, virtual machines and applications. It preserves each report’s input, then gathers fresh evidence for the local model when the incident changes.

PrometheusGrafanaPythonOllamaQEMUGarageOpenEMR
Explore case study
Recorded local lab / 2026-10-26
AlertFresh evidenceRevised investigation
Selected readings before and after the deliberately changed lab fault
Recorded checkFirst reportFresh report
VM can executeNoYes
Storage checkPassedFailed
Application readyFailedFailed

Same failed application check.
Different next investigation.

Operator changed the lab fault · 3/3 live runs verified · human review required

Wireless / Measurement tools

Calibrate Ekahau

Verified sample · live collection not yet tested

I built a Python workflow to compare survey and controller readings by location. It checks which readings can be used and keeps every exclusion in the workbook, so I can see what needs another measurement.

PythonCisco RESTCONFCSVopenpyxlStatistical analysis
Explore case study
Verified synthetic sample / 2026-10-26
Survey pointsController samplesInspectable report
Recorded synthetic comparison summary
Locations compared8 / 13
Locations left out5
Median reading difference-4.00 dB

See what was compared.
See what was left out.

Generated readings · controller − survey · not a calibration correction

AI workspace / Application development

GPTable

In development · authenticated workflow verified

I’m building a workspace where an AI response becomes editable notes, with the source kept alongside them. I can save my own edits and review the next proposal before applying a change.

Next.jsTypeScriptFastAPIPostgreSQLPython
Explore case study
GPTable / Backend workflow · 2026-10-27
RequestManual editReviewed change
6 saved filesSource + 5 notes
2. Design Alternativesv1 → v2 → v3
Reopen saved versionSame saved content

Keep the files.
Keep working.

13 checks · actual UI captures · real sign-in · PostgreSQL · fictional brief

02 / A little context

Look one layer deeper.

I’ve supported Windows systems, compared wireless surveys with controller records and brought network telemetry into monitoring tools. I’m still interested in the same question: what’s really happening here?

Lately, I’ve also been working with local models and building interfaces where I can inspect a suggestion, edit the output and return to saved work.

03 / Say hello

Let’s talk.

Email me about a role, a project, or something you’d like to explore.

Get in touch