Most discourse around AI is hype or doom; most consultancy is implementation-by-the-month or vendor cheerleading. ZeroNonsense.dev does the opposite: build the things, learn what's actually true about AI and the data underneath it, and translate that into advice teams can act on.
Run real projects to challenge assumptions about AI. Every project exists to find what doesn't work as advertised, and what does.
Monetise only where it doesn't compromise the learning. Some projects are infrastructure, some are products, and the difference stays clear.
Sell informed opinion, and a proof of concept where a decision needs one, not delivery. Full implementation contracts overrun, dilute the message, and steal time from the projects.
Every data engagement is framed by what AI and Agentic AI actually need to work: governance, lineage, and quality in service of systems that ship, not data projects for their own sake.
The credibility layer that separates a consultancy from a CV. The interesting part is always what doesn't work as advertised.
A facility-grounded sustainability benchmarking engine on public emission registers: 8 national PRTRs, 4.8 million rows, row-level lineage, running in Azure under a ZeroNonsense.dev IP licence. The hard part is not the pipeline; it is entity resolution. Plants are named after sites, companies after legal entities, and nothing joins for free.
Azure SQL · Python · Public registersThe control plane that runs the company: project state, deadlines and health live in plain git-versioned files with one JSON read-model on top, and every front-end is swappable. The local voice assistant (wake word to tool execution in under 2 seconds on an RTX 4060) is just one surface. Proved portable the hard way: the whole tree changed machines and AI subscriptions in one afternoon, nothing lost.
Files as source of truth · LangGraph voice surface · JSON read-modelMy expertise as versioned, substrate-neutral agent definitions, each with its own self-test. The pilot reviews architecture designs; its first self-test came back perfect and was thrown out on audit, because the harness was leaky. The eval discipline is the product. The honest early signal is review time cut by roughly 84%, with minor rework. I run the same definitions across the harnesses and models below, in the cloud and on an 8 GB laptop GPU: the harness shapes behaviour more than the model does, and what works on GPT often breaks when it moves to a local 7B.
Agent evals · Frozen hold-outs · Markdown + JSON definitionsHarnesses: Claude Code · Claude Cowork · VS Code Agent Mode · Microsoft Copilot · GitHub Copilot · ChatGPT Agent · Hermes AgentA command that runs long development loops (spec, plan, code, test, iterate) with a live dashboard and hard guardrails: agents write and commit locally under their own name, humans push, merge and declare done. Its first real run took a data-pipeline change from spec to merged PR with 73 tests green, overnight.
Python stdlib · NDJSON audit log · Review gatesThe small tools that keep a one-person company honest. BurnRate measures real token spend from session logs (finding: 95 to 98 percent of input was cached context, not fresh work). Voice-driven Time Registration turns spoken sessions into an audit-grade hour log for R&D tax credits. Commands like /log-time-from-sessions glue them together. Also in the drawer: SentinelPR and ReleaseScribe, two open-source GitHub Actions for AI PR review and changelog generation.
Python · JSON stores · SQLCipherA daily AI-accountability content brand. n8n on Hetzner, Claude for copy, HeyGen for video, the Meta Graph API for publishing. Ships every day. The lesson: LLM dedup is harder than it sounds at scale.
n8n · Hetzner · Meta Graph APIAs Director of Architecture and Software Engineering, took a Competitive Market Intelligence platform from a Gmail inbox and a spreadsheet to a six-layer, multi-tenant SaaS product serving hundreds of enterprise customers, engineering team from 2 to 28 over eight years without accumulating legacy. Now Valona Intelligence, a Leader in the 2026 Gartner® Magic Quadrant™ for Competitive and Market Intelligence Platforms. The hard part at that scale is not features; it is keeping a growing system legacy-free while it ships.
Multi-tenant SaaS · Six-layer architecture · Gartner LeaderA solo big-data and IoT initiative at Damen Shipyards became a CIO-sponsored, 30-person program ingesting vessel telemetry for fleet-wide monitoring, and then its own business unit. The hard part was never the sensors; it was making telemetry from hundreds of vessel types mean the same thing before anyone trusted a dashboard built on it.
IoT telemetry · Big data · Fleet monitoringTalks and advisory time from someone who actually builds these systems.
45-60 minute sessions on Agentic AI for engineering audiences. Built from real experiments: local-first agentic stacks, LLM failure modes, and the data foundations agents actually need. Not vendor decks.
Book a talk →Half or full day: architecture review, agent-vs-workflow decisions, data-readiness and governance checks, and an honest AI ROI reality check. You bring the system; I tell you where it breaks and why.
Discuss your project →When a decision hinges on whether something actually works, a small, honest proof of concept: scoped to the risky question, built to be measured, and handed over with the numbers. A PoC to settle the question, not a foot in the door for an implementation contract.
Scope a PoC →SentinelPR and ReleaseScribe: free, open-source GitHub Actions for AI-assisted PR review and changelog generation. MIT and BUSL-1.1 licensed.
View on GitHub →This is for teams deciding what to do about AI and Agentic AI who want a tested, independent opinion instead of a vendor pitch or an implementation contract. You bring the decision. I bring what actually broke, and what held up, when I built the equivalent system myself.
Half a day minimum, on site or remote. Your plans and architecture challenged against what I have actually built and measured. Ends with a short written opinion you can circulate internally.
A 45-60 minute session for engineering audiences, built from real experiments rather than a vendor deck. Includes the specifics: what the numbers were, and where they came from.
An independent read on a vendor, an acquisition target, or an internal build proposal, before budget commits to it. Straight answer on whether the claims survive contact with production.
A second opinion for a CTO or engineering lead before a decision goes to the board: agent versus workflow, build versus buy, and where the data foundation will not hold.
When the decision hinges on whether something actually works, a small PoC scoped to the risky question, built to be measured, and handed over with the numbers. Enough to settle it, not a foot in the door for an implementation contract.
No full implementation contracts, no staffing, no vendor resale. A scoped proof of concept to answer a decision is fair game; a team on the ground for the next six months is not, and I will say so rather than take the retainer. The advice is the product; building it out is someone else's job, or yours.
Measured 85-98% of token spend in long agentic sessions was overhead, then cut it with concrete session-management fixes.
Built a multi-country emissions data pipeline: eight national registers plus the EU-wide E-PRTR, 4.8 million verified rows with row-level lineage.
Run the company on a portable file-based control plane; a voice agent on an 8 GB laptop GPU is one of its surfaces. Moved the whole tree across machines and AI subscriptions in an afternoon, nothing lost.
Cut design-review time by roughly 84% with a self-tested expertise agent, and threw out its perfect self-test score after auditing the test itself.
Let an agent take a data-pipeline change from spec to merged PR overnight, 73 tests green, while humans kept push, merge and declare-done.
Ship a fully automated daily content pipeline that has posted every day for months, and caught its own silent channel outage with a zero-API audit.
As Director of Architecture & Software Engineering at Valona Intelligence, took a Competitive Market Intelligence platform from a Gmail inbox and a spreadsheet to a six-layer, multi-tenant SaaS product serving hundreds of enterprise customers, growing the engineering team from 2 to 28 across eight years without accumulating legacy.
At Damen Shipyards, turned a solo big-data and IoT initiative into a CIO-sponsored, 30-person program ingesting vessel telemetry for fleet-wide monitoring, which became its own business unit, Damen Digital.
As Senior Technical Advisor to HuurPrijsHulp, an Amsterdam start-up, advised on data management, architecture and future-readiness so a small team could automate its internal processes and scale. The same range of advice holds for start-ups and scale-ups, not just enterprises.
The platform I architected, now Valona Intelligence, was named a Leader in the 2026 Gartner® Magic Quadrant™ for Competitive and Market Intelligence Platforms. Under its earlier A-INSIGHTS name it was an FD Gazellen winner (2017 and 2018) and ranked in the Erasmus Top 250 Groeibedrijven (2018 and 2019).
During my advisory work, HuurPrijsHulp was named the third most innovative social-impact company in the Netherlands at the 2024 KVK Innovatie Top 100.
Invited speaker at the 4th Thisworkz meetup (October 2024), the knowledge-sharing event of a six-company Dutch engineering collective: a session on using AI for data platforms, for a cross-company engineering audience. See the event post →
Available for talks and conference sessions on Agentic AI and data foundations. First-hand material and real numbers, no vendor slides. New sessions are added here as they happen.
One person plus agents is a real operating model in 2026. The design question is not what the agent can do, but which verbs you refuse to delegate: push, merge, declare done.
A portability stance, not a privacy one: keep your project's history, decisions, and state in three plain files you own, and any AI tool can pick up exactly where the last one left off.
A practical follow-up: compacting, context trimming, handover files, and model routing to reduce token overhead in long agentic sessions.
Advisory enquiries, talk invitations, or questions about the tools. I read everything.
Connect on LinkedIn