ML Engineer

Kaden VanHoecke

ML Engineer focused on computer vision, real-time systems, and production ML pipelines. Background in full-stack development, automation, and neural networks. Targeting BCI/AI roles at frontier technology companies.

Get in touch View work

About

I build AI systems that run in production, not in notebooks. My work sits at the intersection of ML engineering, agentic AI, and industrial automation - turning manual processes into autonomous pipelines and embedding intelligence directly into the tools people already use.

McQueenyML is a 74K LOC platform I built and operate for an HVAC sales engineering team. It cut estimating workflows from hours to minutes with a 5-tier matching engine (pgvector embeddings, trigram fuzzy search, Claude AI fallback) across 553K indexed files and 3,000+ equipment orders synced from SharePoint. Adze is a native COM add-in that brings AI into SOLIDWORKS as a first-class citizen - 19 typed grounding tools reading live CAD sessions, with an 8-step governed safety lifecycle for write operations. Project Ava is a self-evolving AI ecosystem on dedicated hardware: persistent memory (brain.db), orchestrated domain agents, spaced-repetition learning, automated security auditing, and a document crawling pipeline - all accessed via mobile-first PWA over Tailscale mesh.

Mechanical engineering background (KSU), CSWA certified. Everything else is self-taught and production-proven.

Featured Projects

Project Ava

github

Dedicated headless server running as a persistent, self-evolving AI assistant. brain.db provides structured memory across 130+ sessions. 22 domain-specific agents handle tasks from tax compliance to 3D print calibration. Mobile-first PWA serves 15 feature tabs with inline editing, real-time monitoring, and Mercury banking integration. Express 5 API with Zod validation, ChromaDB semantic search, and OpenClaw agent orchestration — all accessed over Tailscale mesh from any device.

Persistent AI ecosystem: 22 orchestrated agents, structured memory (brain.db), 262 smoke-tested endpoints, mobile-first PWA over Tailscale mesh

React 19Vite 8Tailwind 3.4Express 5TypeScript 5.9SQLite (better-sqlite3)ChromaDBOpenClaw

CloudBooks

live

SaaS accounting platform deployed on Cloudflare Pages with Supabase backend. Multi-tenant architecture with row-level security, Stripe payment integration, and real-time data sync across devices. Automated transaction categorization using pattern matching. Built on React, TypeScript, Supabase (PostgreSQL + Auth + Edge Functions), and Cloudflare Workers.

Production multi-tenant SaaS with Stripe billing, real-time sync, and automated transaction categorization

React 18, TypeScript, Tailwind CSS, Vite, Zustand, TanStack Query, React Router v6, Supabase AuthPostgreSQL, Vitest, Playwright, pg_cron

McQueenyML

Production automation platform for an HVAC contractor's sales engineering workflow. 5-tier matching engine (pgvector embeddings, trigram fuzzy search, Claude AI fallback) indexes 553K files and 3,000+ equipment orders synced from SharePoint. Processes 10K+ XML specifications with 100% parsing accuracy. Built on Next.js, FastAPI, PostgreSQL, and ChromaDB with real-time sync pipelines.

4 hours to 12 minutes: production ML platform serving a sales engineering team across 553K indexed files

Frontend: ReactTypeScriptViteTailwind CSS. Backend: Python 3.11FastAPIuvicorn. Database: PostgreSQL via Supabase (pg_trgmpgvector). ML: Ollama (qwen3:14b, nomic-embed-text). Infrastructure: 7 NSSM services on Windows 11, Cloudflare Tunnel.

Adze — AI for SOLIDWORKS

github

AI design assistant that runs inside SOLIDWORKS as a COM add-in, not alongside it. 19 typed grounding tools read live CAD sessions (dimensions, mates, mass properties, BOM). 8-step governed safety lifecycle for write operations ensures AI modifications are reversible and auditable. Built in C# on .NET with the SOLIDWORKS API. SOLIDWORKS Solution Partner Program application in progress.

Native in-process AI assistant for SOLIDWORKS with 19 typed grounding tools and governed write safety

C# .NET 4.8, SOLIDWORKS COM API, NUnit 3, MSBuild, PowerShell, WinForms WebBrowser

Oracle (Trade Signal)

Self-directed market analysis system that generates hypotheses, tests them against historical data, and manages paper portfolios with real risk controls. 4 ML cortexes (LSTM prediction, sentiment analysis, technical patterns, anomaly detection) feed an Oracle decision engine with 8-dimension risk scoring. Walk-forward validation prevents lookahead bias. Built on Python, PyTorch, and FastAPI with real-time data ingestion.

Autonomous curiosity engine: 4 ML cortexes, 8 risk dimensions, walk-forward validated paper trading

Python 3.13 (miniconda3), FastAPI, SQLite, TensorFlow, PyTorch, transformers (FinBERT), React/TypeScript/Tailwind (ava_hub UI)

Skills

Languages

Python TypeScript C# C/C++ Rust SQL

Frameworks

React Node.js Express FastAPI .NET / COM Interop PyTorch YOLO TensorFlow

Tools

Docker n8n SQLite PostgreSQL Supabase ChromaDB Git / GitHub Tailscale Vite SOLIDWORKS

Domains

Computer Vision Neural Networks Agentic AI Systems Automation Real-Time Systems CAD / Mechanical Engineering

Strengths

Problem Solving Self-Directed Learning Technical Writing

Experience

Founder & ML Engineer

VHTech LLC - Lenexa, Kansas

Jan 2023 - Present
  • Founded ML engineering consultancy (LLC formed March 2026, freelance since 2023)
  • Built and operate McQueenyML for McQueeny Group Inc. - 74K LOC HVAC estimating platform with 5-tier ML matching engine (pgvector, trigram, Claude AI fallback) serving a sales engineering team in production
  • Building Adze - native COM add-in embedding AI directly into SOLIDWORKS with 19 typed grounding tools and an 8-step governed safety lifecycle for write operations. SOLIDWORKS Solution Partner Program application in progress
  • Built Project Ava - persistent AI ecosystem on dedicated hardware with orchestrated domain agents, structured memory (brain.db), automated security auditing, and document crawling. Mobile-first PWA over Tailscale mesh
  • Built CloudBooks - SaaS accounting platform on Cloudflare Pages with Supabase backend, real-time sync, and automated transaction categorization
  • Designed n8n automation pipelines that cut document processing from 4 hours to 12 minutes across 10K+ XML specifications

ML Engineer (Contract)

McQueeny Group Inc. - Texas (Remote)

Dec 2025 - Present
  • Sole developer on McQueenyML - production HVAC estimating automation deployed to a sales engineering team
  • Built a 5-tier product matching pipeline using pgvector embeddings, PostgreSQL trigram search, and Claude AI fallback with a self-improving corrections loop
  • Automated SharePoint sync across 372+ active projects - 3,000+ equipment orders ingested via Microsoft Graph API
  • Operate a 7-service Windows production stack (API, UI, Cloudflare tunnel, filesystem watcher, intelligence daemon, log shipping) via NSSM
  • Manage 553K-row file index across network drives with real-time project lifecycle tracking

Sprayers Technician

Wallboard Specialties - Lenexa, KS

Jun 2018 - May 2025
  • Ceiling texture spraying crew - summers 2018-2023, full-time 2023-2025
  • Equipment operation, material mixing (215-gallon batches), job site coordination across commercial projects
  • 7 years of physical trade work while simultaneously self-teaching software engineering and ML

Freelance Developer & Automation Specialist

Self-Employed - Texas

Jan 2023 - Present
  • Built and deployed CloudBooks SaaS accounting platform serving production users
  • Designed n8n automation pipelines reducing document processing time by 95%
  • Developed Project Ava: self-evolving AI assistant with 193 API endpoints and 368 tests
  • Created real-time ML systems including YOLO-based computer vision applications

Mechanical Engineering Coursework

Kansas State University - Manhattan, Kansas

Aug 2021 - May 2023

Contact

kaden@vhtech.me

Lenexa, Kansas