Years of pattern recognition
From 8-bit BASIC to distributed systems, product code, data tools, and modern AI-assisted development.
Programming since Apple ][+ BASIC. Shipping with AI now.
Old-school software judgment. New-school machine speed.
I help small teams turn AI coding tools into production code: fast prototypes, real releases, fewer theatrics, and code that can survive Monday morning.
The unfair stack
AI coding is only impressive when it ships.
The trick is not coaxing an LLM into writing files. The trick is knowing what to ask for, what to delete, what to test, and what should never have been generated in the first place.
Lines a week, with restraint
High-volume output matters only when it passes review, lands cleanly, and makes the team faster after the merge.
Drama required
Low-maintenance, collaborative, production-minded help for founders and engineering teams that need leverage.
Clickable proof
Recent artifacts, not just claims.
A few public trails from the current workbench: AI evaluation, MCP usability testing, network telemetry, browser automation, embedded devices, local LLM infrastructure, and writing about the strange new practice of pairing with machines.
GitHub / AI evaluation
Which LLM should you trust with agentic work?
Utilities for comparing one-shot prompts across LLMs and agent CLI tools.
View repository GitHub / MCP usabilityWhen bots use MCP tools, how do you observe and reduce friction?
A dual-agent harness where one model uses the tools and another watches for friction.
View repository GitHub / Browser automationWhat can a wandering browser reveal about a web app?
A headed Playwright crawler for first-pass UI exploration and practical web inspection.
View repository GitHub / Rust network scannerWhat is really happening on your home network?
Rust packet capture, Nmap device enrichment, SQLite persistence, and a Leptos dashboard.
View repository GitHub / Local AI stackHow fast can you push a brand-new Intel Arc Pro B70 GPU?
Local LLM stack on Intel Arc Pro B70: llama.cpp SYCL, llama-swap, opencode, and pi.
View repository GitHub / Embedded Telegram botCan a tiny $15 board help people check in on each other?
A Raspberry Pi Pico W wellness bot that pings people and alerts a buddy if they miss.
View repository LinkedInWhat did Jeff learn before the bots showed up?
Articles, career context, and older receipts that live better on the professional timeline.
Open LinkedIn BlogWhat did Jeff learn after the bots showed up?
Notes from the front line of AI-assisted development, judgment, workflow, and taste.
Read the blogWhat this actually looks like
A senior engineer with an absurdly fast pair programmer.
I use AI to fan out implementation, tests, refactors, data wrangling, UI passes, and debugging. Then I bring the grown-up parts: architecture, taste, sequencing, review, and the nerve to say no.
Ways to use me
Not a guru. A force multiplier with commit history.
AI workflow audit
Find where AI can speed your team up without flooding your repo with expensive confetti.
Prototype-to-production sprint
Turn a messy idea into working software, then harden the parts that need to last.
Fractional senior builder
Add practical AI coding horsepower plus decades of engineering judgment, without a full-time hire.
Legacy modernization
Use AI to map, refactor, test, and retire old code carefully, with someone who remembers why it exists.
The ask
Bring me a real bottleneck.
A feature stuck in planning, a codebase nobody wants to touch, a team curious about AI but allergic to hype. I will help you make it move.