Application for AI Product Builder

20+ Years of experience helping businesses leverage technology systems to improve processes. I've owned the problem, the requirements, and the solution design, what I've found recently is that with AI, I can own the build as well, and I love it.

I'm a process improvement expert who spent 15 years designing and implementing enterprise systems, then 5 years managing product at ServiceTitan through their $300M-to-$10B growth phase. What I learned: I'm at my best when product judgment and building aren't separated.

Since December 2025, AI tools have let me do exactly that. I've shipped 20+ production systems — including a multi-workflow automation suite on n8n for a paying CRE client, an AI Chief of Staff with 17 Skills and persistent memory, and an autonomous daily briefing pipeline that runs a local LLM for email privacy. The magic of AI products lives in the space between deterministic workflows, AI agents, and human judgment. I obsess over knowing what belongs where.

n8n Power User Claude Code Daily Driver Local LLM Infrastructure Production AI Systems

Why n8n, specifically

I already live in the product

I run two self-hosted n8n instances (personal + client), built automated provisioning to spin up more, and have shipped production workflow suites using the n8n Workflow SDK. This isn't a demo screenshot — it's my daily infrastructure.

Process improvement is my through-line

From enterprise Salesforce implementations to ServiceTitan product management to AI automation — every chapter has been about making processes better. n8n is a process improvement platform. The alignment isn't coincidental.

Builder energy, product judgment

I've done the PM role. I know how to talk to customers, size markets, and navigate trade-offs. But what I've learned about myself is that I'm at my best when I can combine that judgment with actually building. That's what the builder track is for.

This work doesn't feel like work

Building process automation with AI is the first thing in my career that I can do for hours without noticing. The energy is genuine, the output is real, and the portfolio below is what happens when someone finds the thing they were meant to build.

Workstream Alignment

PRIMARY FIT

AI Building / Super Agent

Skills: 17 production Skills in MARVIN + CRE Field Mapping Skill
Custom Memory: Three-layer persistent state across sessions and machines
Browser Use: Puppeteer automation orchestrated by n8n workflows
Chat-First Building: Meal Planning, MARVIN — conversational input, automatic output
Proactive Suggestions: Daily briefing pipeline, deadline alerts, priority proposals

STRONG FIT

AI Trust

Evals: Burrfect AI enhancement harness, BTC backtesting with statistical validation
Guardrails: CRE Skill's 15 non-negotiable rules, Woordjes rejection system with rate limiting
Reliability: Overnight pipeline runs daily since March 2026 — deterministic + LLM hybrid
Human-in-the-Loop: Interactive briefing checkboxes, approval gates, confirmation flows
Privacy: Local LLM for email processing — sensitive data never leaves the hardware

The Acceleration Curve

Same person. Better tools. The projects got more ambitious while the timelines got shorter.

Oct 2023

Kumpel.ai — RAG Chat Platform

Django LlamaIndex OpenAI

~2 months — Full RAG pipeline before it was common

Oct 2025

BTC Backtesting Engine

Python Claude Statistical Validation

8 days — Eval pipeline with 5 significance tests

Nov 2025

claude-slack

Python Slack SDK WebSocket

3 days (core) — Shared publicly, positive reception

Dec 2025

AI Sea Change

Claude Code Cursor AI-Native

AI tools hit the threshold — started building directly

Jan 2026

MARVIN — AI Chief of Staff

Claude Code Skills MCP

ongoing, daily use — 17 Skills, persistent memory, autonomous agents

Feb 2026

Woordjes

Django Claude API Appliku

4 days → production — AI enrichment + guardrails

Mar 2026

Overnight Briefing Pipeline

Python Qwen 35B MCP

iterative, runs daily — Local LLM for email privacy

Mar 2026

MBO n8n Workflow Suite

n8n Puppeteer Django

ongoing engagement — Production n8n for paying clients

Apr 2026

Burrfect Water

Python SCA Standards

5 days — Deterministic engine, 44 cities

May 2026

Meal Planning Assistant

Claude Code HTML

2 days — Conversational → automatic output

Featured Projects

Ordered by relevance to n8n's AI Product Builder role

MBO Listing Sync — n8n Workflow Suite

Live — Paying Client

A production multi-workflow automation system built on n8n for a commercial real estate client. Webhook triggers, browser automation via Puppeteer, content generation, queue management, and a thin-loader architecture where fixes deploy to all client instances without per-instance updates.

n8n relevance: I don't just use n8n — I build production workflow suites on it for paying clients, and I deploy self-hosted n8n instances with automated provisioning.

n8n n8n Workflow SDK Puppeteer Webhooks MCP

Read the full case study →

MARVIN — AI Chief of Staff

Live — Daily Use

An AI system with persistent memory, 17 Skills, autonomous agents, and a daily briefing pipeline. Extended from an open-source framework into my primary work orchestration tool — running daily across two machines since January 2026.

n8n relevance: Skills, custom memory, autonomous agents, proactive suggestions — MARVIN's architecture maps directly to n8n's AI Building / Super Agent workstream.

Claude Code Skills Custom Memory MCP Autonomous Agents

Overnight Briefing Pipeline

Live — Runs Daily

A local LLM reads my email at 5 AM, translates the Dutch ones, cross-references my task list, reconciles completions, and publishes an interactive HTML briefing with checkboxes I can tap from my phone — all without sending email content to any cloud provider.

n8n relevance: Hybrid deterministic + LLM architecture, local LLM for privacy, guardrails, multi-tool orchestration, human-in-the-loop — the "AI works in production" story.

Python Qwen 35B (local) MCP Deterministic + AI Hybrid Guardrails

Read the full case study →

CRE Browser Automation

Live

Shared Puppeteer scripts for syncing commercial real estate listings to CoStar, Crexi, and Catylist. Thin-loader architecture where n8n Code nodes fetch scripts at runtime from GitHub — fix a selector once, all client instances pick it up.

n8n relevance: n8n's AI Building workstream specifically lists "computer and browser use." This is production browser automation orchestrated by n8n.

Puppeteer n8n Browser Automation Anti-Detection Error Taxonomy

Burrfect AI Enhancement Pipeline

Live

Built the AI catalog enrichment system for a live consumer app — originally on Google Apps Script calling Perplexity for search + OpenAI for structured JSON (before either had native support), then migrated to Cloud Functions with cold-start management and dependency-chain orchestration.

n8n relevance: Built the eval/testing harness for regression testing model upgrades against 20+ records. The "measure before you trust" mindset n8n's AI Trust workstream requires.

Perplexity AI OpenAI Firebase Cloud Functions Eval Harness

Read the full case study →

CRE Field Mapping Skill

Live

A structured "Skill" (n8n's exact term) that guides AI through onboarding new CRE platforms — screenshot audit, JSON field map creation, DOM inspection, Puppeteer script building, and E2E testing. 15 non-negotiable rules encoded from production failures.

n8n relevance: This IS a Skill in n8n's vocabulary — making AI reliable at complex multi-step tasks. The exact gap n8n's Skills feature aims to close.

Claude Code Skill Structured AI Guardrails Human-in-the-Loop

More Projects

Additional evidence of builder depth and range

n8n Deployment Infrastructure

Live

Automated client provisioning for self-hosted n8n instances on Appliku. provision.py spins up isolated instances with PostgreSQL in ~2 minutes.

DockerPythonn8nPostgreSQL

Personal n8n Instance

Live

My own n8n at n8ndan.applikuapp.com — Blitzit task filtering, Gmail pre-filtering, and Cowork data pulling. Tailscale-in-Docker for reaching local LLMs from the cloud.

n8nDockerTailscaleLocal LLM

Meal Planning Assistant

Live

A short conversation to plan 3 family meals, then it auto-generates shopping lists organized by my grocery store's layout and recipe walkthroughs with built-in timers. Published to a mobile-accessible HTML server.

Claude CodeHTMLTailscale

Woordjes — Dutch Learning App

Live

AI-enriched vocabulary flashcards with dual LLM backend (cloud + local), guardrail rejection system with rate limiting, and background processing queue. Built for myself as an American learning Dutch.

DjangoClaude APIGuardrailsHTMX

Case study →

Local LLM Infrastructure

Live

Running 20+ models on a MacBook Pro M2 Max — llama.cpp, MLX, OpenAI-compatible proxy with SSE stream coalescing, Tailscale remote access. The practical constraints of model serving, quantization, and latency.

llama.cppMLXFastAPITailscale

BTC Backtesting Engine

In Development

Data pipeline with statistical validation (5 significance tests per backtest), walk-forward overfitting detection, and regime-aware evaluation. The backtesting system IS an eval pipeline.

PythonpandasStatistical Validation

Case study →

Background

2025 —

AI Product Builder — Consulting & Personal Systems

Building AI-powered automation for consulting clients (MBO — commercial real estate) and shipping personal production systems. n8n workflow suites, Django apps, browser automation, local LLM infrastructure, autonomous agent pipelines.

2022 —

Co-Founder — Burrfect

Consumer espresso app (iOS + Android). Built the AI catalog enhancement pipeline, eval harness, and analytics dashboard. PM'd the 12-model recommendation engine with the engineering team.

2017 — 2022

Product Manager — ServiceTitan

New market strategy during the $300M → $10B growth phase (now public on NASDAQ). Market sizing, competitive analysis frameworks, and cross-team coordination across six product teams evaluating TAM-doubling expansion opportunities.

2011 — 2017

Solution Architect → Product Manager — ServiceMax

Enterprise field service management on Salesforce. Progressed from professional services consultant to partner enablement architect to product management. Schedule optimization, dispatch, and internal tooling.

1999 — 2003

B.A. Computer Science — Macalester College

St. Paul, Minnesota

Let's build together

Based in Haarlem, Netherlands — 20 minutes from Amsterdam by train. US citizen, available for hybrid or remote. Ready to start immediately.

LinkedIn Profile → dantana@gmail.com