AI agents, RAG pipelines, MCP servers, and web/iOS products — shaped by evaluations, edge-case testing, and the quality gates that separate a reliable system from a good demo.
Live
IFTA agent in production
429
Tests on the IFTA pipeline
2
MCP servers, self-hosted
Evals
RAG graded by a benchmark
Core Capabilities
AI automation, full-stack products, quality engineering, and local Sacramento computer support — built around real outcomes.
AI agents, RAG systems, prompt engineering, workflow automation, model evaluations, fine-tuning, and business process automation.
Starting at
Intro -20%$8,000$6,400intro project · $150/hr advisory
Modern websites, web applications, dashboards, APIs, iOS apps, MVP builds, and production-ready digital products.
Websites from
Intro -20%$1,500$1,200websites · apps/MVPs $7,500 → $6,000
SDLC/STLC-based testing, functional testing, regression testing, smoke/sanity testing, test design techniques, edge-case validation, and release confidence.
Starting at
Intro -20%$3,000$2,400intro project · $90/hr after scope
Diagnostics, setup, upgrades, cleanup, and practical troubleshooting for local customers in the Sacramento, CA area.
Flat rate from
Intro -20%$49$39first diagnostic · Sacramento
Most AI projects fail in the gap between prototype and production — great in the demo, brittle in real use. I close that gap by treating AI engineering, full-stack development, and software quality as one build path instead of three separate disciplines.
Agents, RAG pipelines, and web/iOS products designed with evaluations, edge-case coverage, and regression checks from day one — so what ships behaves the way you expect when real users start poking at it.
Behind ArtJeck is one engineer with a software-quality background — so testing isn't a phase bolted on at the end, it's how the system gets built.
AI workflows shaped around business outcomes
Full-stack products with maintainability in mind
iOS experiences designed for focused mobile use
Testing strategies that improve release confidence
Six stages that keep product decisions, technical execution, and quality moving together.
Clarify the business goal, users, constraints, and what success needs to look like.
Design the product flow, data model, AI approach, integrations, and testing strategy.
Develop the automation, web platform, API, or iOS product with maintainable code.
Validate functionality, edge cases, regressions, AI outputs, and release readiness.
Ship to production with clear configuration, monitoring paths, and handoff notes.
Use feedback, evaluations, and product data to refine the system after launch.
Production sites, AI agents, MCP servers, and the live ArtJeck platform — shipped, documented, and open to inspect.
Trucking carriers spend hours every quarter reconciling fuel and mileage data, hand-typing per-state lines into the gov portal, and second-guessing whether the math matches what the state will recompute.
End-to-end pipeline that ingests raw mileage and fuel files, computes a state-portal-ready return with exact CDTFA math, and runs an AI agent over it to flag missing surcharges, MPG anomalies, and audit-bait fuel patterns before filing.
Python 3.12, pandas, openpyxl, pdfplumber, Anthropic Claude Opus 4.7, multi-tenant client registry, per-truck Excel deliverables, CLI + agent tools
429 automated tests including a real-data backtest that matches a Kentucky carrier's Q4 2025 CDTFA filing to the penny. Per-truck reconciliation tested to fleet totals within $5 rounding drift.
First active client (DM Express Inc., KY) filing quarterly through the pipeline. AI agent produces a structured review note with concrete next-steps before each filing.
Cloud AI assistants send your notes, docs, and client files to someone else's servers, and answer without showing where the answer came from — a non-starter for sensitive or auditable work.
A local-first RAG assistant that ingests your own notes, docs, and code and answers only from them, with every claim cited to its source. Free local models by default through an OpenAI-compatible gateway; one env flag routes the answer to Claude.
Python, OpenAI-compatible LLM API, Qdrant / Chroma vectors, LangGraph workflow, SQLite task state, MCP server (stdio + token-gated HTTP)
Eval harness over a synthetic regression corpus — retrieval hit-rate, grounded-answer checks, and abstention cases — plus a pytest suite.
Open source and in daily use; exposed over MCP so Claude Desktop and Claude Code can query and teach it directly.
Operating a multi-machine AI lab from anywhere usually means SSH and a raw shell — powerful, but reckless to hand to an autonomous agent.
An MCP server that gives any agent safe “hands” on the lab: health checks, model management, free local inference, and a deliberately gated remote shell — allowlist only, no shell metacharacters, hard timeouts.
Python, MCP (stdio + token-gated HTTP), httpx, subprocess argv (no shell=True), launchd, Tailscale
pytest suite covering the command-gating safety logic, runnable offline.
Open source and running 24/7; lets an agent operate the lab from an iPad while the dangerous operations stay locked behind validated tools.
Daily inbox review was noisy and manual, with important messages mixed into low-priority mail.
A local agent pulls Yahoo mail over IMAP, uses a local Ollama model to triage and summarize, then sends a Telegram digest.
TypeScript, Node 20, imapflow, mailparser, Ollama, SQLite, Telegram Bot API, launchd
Includes test-connection and dry-run paths, SQLite dedupe for idempotency, and cost reporting per run.
A self-hosted morning digest that runs on schedule, entirely on local Ollama models at $0 API cost.
A small jewelry brand needed a real online store — catalog, reviews, promotions, and self-service management — without a monthly platform fee or a dashboard to learn.
A full Next.js storefront with a product catalog, image galleries, reviews, stock, a Buy-Now flow, and token-based personal offer links — plus a secured admin where the owner manages products, images, and offers.
Next.js 16, React 19, TypeScript, Prisma + libSQL (Turso), Vercel Blob image uploads, Tailwind CSS v4, TOTP two-factor admin
Admin gated behind TOTP two-factor auth, validated image uploads, and cached DB queries for fast product pages; storefront and admin flows tested end-to-end by hand.
A live storefront the owner runs themselves — add products, upload images, publish reviews, and send personal offer links — with no recurring platform cost.
Bills of lading are still often reviewed and keyed manually, which slows operations and creates avoidable data-entry risk.
An in-progress extractor for turning BOL documents into structured shipment data with validation and review-ready output.
OCR/document parsing, LLM structured extraction, JSON schema validation, field checks, export workflow
Planned around sample BOL fixtures, required-field checks, edge-case documents, and regression tests for extraction quality.
In progress: designed to reduce manual BOL entry and make shipment data easier to review, reuse, and automate.
A small trucking company needed a credible, fast, phone-friendly site to help recruit drivers.
A one-page React/Vite site with light/dark theme, animated sections, and a structured driver application flow.
React, TypeScript, Vite, Vitest, plain CSS, mailto application flow, Vercel
15 tests using equivalence partitioning, boundary value analysis, decision tables, and state transition testing.
A sanitized real-client portfolio build tuned for iPhone behavior, accessibility, and sub-500KB first load.
The brand needed a live professional site that clearly positions AI automation, product engineering, and QA discipline.
A production Next.js portfolio with clean dark UI, glass-card components, scroll animations, contact form, and full SEO metadata.
Next.js, React, TypeScript, Tailwind CSS v4, Resend-ready contact action, Cloudflare email routing
Verified with lint, production build, browser checks, contact email routing, and an audit-clean dependency override.
A live personal brand site at artjeck.com with direct email contact and a deployable project showcase.
second-brain and lab-control (in Work above) don't run in the cloud — they run on a private, always-on stack I operate myself: free local models by default, paid cloud only when it earns it.
Every model call goes through a LiteLLM gateway — free local models by default, automatic failover between machines, and a hard $50 / 30-day cap before anything paid runs.
Two always-on machines run Ollama for free local inference — chat, code, and embeddings — with Qdrant holding the vector index and Postgres the structured state.
Everything sits on a private Tailscale mesh — nothing public. The MCP servers are exposed over token-gated HTTP and kept alive around the clock with launchd.
Free-local-first by default; nothing is exposed to the public internet.
The skill set is cross-functional so implementation, testing, and launch readiness stay connected.
Bring the workflow, product, AI idea, or Sacramento-area computer issue you want to move forward. For repair requests, include your city, device, and the problem you're seeing.
Best Fit