Innovation

Where I’m headed

I believe the next decade of software belongs to teams that build with AI as a first-class engineering method — and to products where AI is the interface, not a bolt-on. These are the directions I’m building toward at Clever Initiative.

Visit cleverinit.com (opens in a new tab)
Proven

A filed patent

I don’t only execute on the state of the art; I help define it.

Patent · Netherlands · 2025

An AI-driven healthcare system

A patent filed in the Netherlands (2025) for an AI-driven healthcare system. The specifics are protected; the principle is the one behind everything I build: intelligence designed into the core, not added at the edges.

R&D directions

Four ideas I’m incubating

The four directions I lead R&D on at Clever Initiative — each shipping as a CleverInit module as it matures. Described here by the problem they solve and the principle behind them. All four are in development, not released.

  • R&D 01In development

    AI Copilot for business operations

    Problem
    Operators run SaaS platforms through dense dashboards and support tickets — slow, manual, and easy to get wrong.
    Approach
    An AI-native, tool-using assistant that runs the platform in plain language — “draft this invoice”, “who hasn’t paid in 30 days?” — through auditable, permission-checked tool-calls, with per-organization provider choice and hard cost controls.
    Edge
    Every consequential action is an audited, permission-checked operation — built for regulated industries, not a chatbot in an iframe.
  • R&D 02In development

    Embedded SaaS analytics

    Problem
    Most platforms sit on a goldmine of event data but ship dashboards nobody reads — and a separate BI tool costs a fortune.
    Approach
    An opinionated, SaaS-aware analytics layer that turns a platform’s own event stream into executive dashboards (MRR, churn, DSO), scheduled reports, anomaly alerts, and plain-language narratives.
    Edge
    It collapses a €25k/year BI tool into something an operator actually reads — and it understands SaaS out of the box.
  • R&D 03In development

    No-code workflow automation

    Problem
    Businesses pay for three external automation subscriptions just to move data between the tools they already use.
    Approach
    A visual, durable automation engine native to a platform’s own data and events — triggers, actions, branches, loops, human approvals, and AI steps, with durable execution and versioning.
    Edge
    One built-in engine replaces the Zapier/n8n sprawl, with first-class access to the data it’s automating.
  • R&D 04In development

    AI-driven product intake — “Forge”

    Problem
    The pre-sales funnel for custom software is an 8–20-hour grind of calls, scoping, and proposals before anyone builds a thing.
    Approach
    A self-serve portal where a founder talks to an AI agent and leaves with a generated business plan, pitch deck, and a real, scoped implementation plan.
    Edge
    It collapses that funnel to a 30–60-minute review — and auto-generates handover tutorials on delivery.

The thread that connects them

Each of these is AI-native, auditable, and built for real businesses — not demos. Together they describe one way of building software: intelligent by default, honest about what it’s doing, and accountable for every action it takes. It’s the same standard I bring to how I work.

See how I build
Perspectives

How I think about AI and the craft

Short posts I’ve shared on LinkedIn about AI, architecture, and what it now takes to be a software engineer.

  • LinkedIn post

    “It depends.”

    Architecture is trade-offs. I’d rather understand the problem, the constraints, and the team than chase a definitive answer.

    It’s the most grown-up answer a software engineer can give. … For what problem? Under what constraints? For which team? … Maybe that’s the shift that slowly turns an engineer into an architect. Chasing definitive answers less, and understanding the problem more.

    Find it on LinkedIn
    • Architecture
    • Trade-offs
    • AI engineering
    • Hiring

    LinkedIn post

    This is not made by AI!

    A developer interview should observe how someone works with AI: the first prompt, breaking a goal into phases that fit a context window, reviewing and testing AI-generated code, spotting anti-patterns, building a harness around an agent, OWASP-minded security thinking, and token-usage discipline.

    • AI
    • Leadership

    LinkedIn post

    AI can write code.

    AI writes code, but the human skills around it — clients, judgement, leadership — are what make an engineer. Combine them with AI and you have the future of the industry.

    • AI
    • Future of software

    LinkedIn post

    We are the last generation of programmers!

    The programmers of the future will prompt rather than program — and one day they may look at the Linux kernel the way we look at the Pyramids.

Let’s build

Building something in this space, or want to?

If this is where you’re headed too, I’d love to compare notes — or build it with you.

Let’s talkView résumé