Independent engineering consultancy · Europe / North Africa · Since 2025

From vibe coding
to real product.
In 90 days.

Engineering maturity for AI products. Daily releases, sovereign AI, zero slop.

Evals Rollback Observability Domain model Any model per feature

01 · What I do

Three things. Every engagement.

01 / Engineering

Engineering uplift

I bring your AI product up to the engineering bar your customers actually pay for: evals, rollback, observability, a clean domain model, and a CI/CD pipeline that ships multiple releases a day. TDD, DDD, XP — applied to AI features, not abandoned because of them.

02 / Infrastructure

Sovereign AI stack

I install model-agnostic AI infrastructure you actually own. Run OpenAI, Claude, Gemini, Grok, or local open-weight models — scoped per feature, swap anytime. No vendor lock-in, no data leakage, no surprise pricing.

03 / Leverage

AI as internal factory

I wire AI into your engine room before your product. Docs, demo videos, refactors, test generation, metrics analysis, code review. Your team stops drowning in busywork and reclaims the time that AI can't give them back — time with your customers.

02 · The bet

Stop flooding users
with AI slop.

The market narrative says: "ship more AI features, because AI." Your users can't keep up. Your product gets unstable. Your trust metrics crater. Meanwhile the frontier model providers profit handsomely from the token-grinding that funds it.

I bet the opposite. AI belongs inside your engine room first: documentation, demo videos, refactors, test generation, metrics analysis, code review — the boring internal work where AI is genuinely good today. That buys your team the one thing AI can't give them: time to talk to the people who actually pay you.

Users feel the difference. Fewer features, better shipped, responsive to feedback. Less slop, more product.

1+/day

Release cadence

What your team ships once the engineering is right. No freeze windows, no big-bang merges.

Any model

Per feature

OpenAI · Claude · Gemini · Grok · local open-weight. Scoped per feature, swappable anytime.

~90 days

Vibe coding → real product

End-to-end engagement window: audit, install, teach, hand back. One part-time engineer left on the other side.

03 · The thesis

By 2028, most B2B SaaS
will be 90% autonomous.

Machines ingesting user feedback, shipping aligned changes. The humans left on the team won't write code — they'll validate results, review designs, approve major ADRs, own SLAs. In one word: they'll be accountable.

The teams that reach that future aren't the ones stacking more tokens today. They're the ones installing the boring engineering that lets software evolve safely on its own. That's the playbook I'm building with every engagement.

Read the manifesto

04 · Who you're hiring

Osmane Kalache.

15+ years shipping real software across IoT, telecom, real estate, and fintech. I started ioprodz in late 2025 to be the counter-voice to AI slop — the one that says the boring engineering is the competitive advantage, and that a small team with the right practices outlasts a big team stacking tokens.

"AI handles the code. You get your time back — for the thing AI can't do: building the relationships with the people who pay you."

DDD TDD XP CI/CD Conway's Law Fast feedback Platform engineering DX

Read the manifesto LinkedIn GitHub Twitter

05 · Let's talk

One conversation
changes the trajectory.

A tailored 30-minute screening. I'll tell you honestly whether your AI product has a vibe-coding problem or an engineering problem — and what the 90-day path looks like if you want to fix it.

Book a tailored screening [email protected]

From vibe coding to real product. In 90 days.