Built for the gap between idea and live in production. A gap most AI shops bill you to discover, not to cross.
Two years building AI in production taught one thing: most strategy work dies in the gap between the deck and live in production. The team that wrote the recommendations isn't the team that has to live with them, and by month three the document is on someone's hard drive, slowly composting.
innops started in 2025 with a single rule: every engagement ends with something running in production. Not a roadmap, not a maturity model. A workflow your team is using on Monday. Strategy that doesn't ship isn't strategy. It's a PDF.
We're AI-native by operation because the work itself is. Research, synthesis, code, evals. The studio runs on AI, which is why we stay small, charge flat fees, and still ship in weeks instead of quarters. Senior engineering at small-studio prices, because we run what we sell.
In that order. When two of them conflict, the earlier one wins.
If a workflow can't reach production in twelve weeks, we're solving the wrong problem. We re-scope before we extend.
The person who sold you the engagement is the person doing the work. No bait-and-switch to juniors. No partner-on-the-cover, associates-in-the-trenches.
Documentation, runbooks, training, monitoring. Your team owns it after we leave. That's the engagement, not the optional Phase 4.
We don't quote "up to 10× productivity". We name the workflow, the baseline, the delta, and the measurement window. If we can't measure it, we don't claim it.
Flat fees per engagement, scoped before we start. Continuing work runs on outcome-based or scope-defined retainers: capped, narrow, terminable monthly. No time-based or activity-based retainers. We'd rather lose the renewal than bill for hours instead of outcomes.
Zakaria leads every engagement himself. A network of vetted freelancers steps in when a project needs specialist hands: retrieval, infra, frontline ops. The lead contact stays the same.
Two years shipping production AI in real businesses: computer vision, MLOps, the unglamorous parts. Now applying the same engineering rigor to operators wanting to move from prototype to production: assistants, automation, agents. Started innops in 2025 to build the AI he kept wishing existed at his last job.
Reads more about practical AI than is probably healthy. Based in Antwerp. Speaks NL / EN / FR / AR.
Useful when you're trying to figure out whether we'll get on. If three of the 'don't' items sound like Tuesday at your shop, we might not be the right fit. Honestly.
A 30-minute call is enough to figure it out. No pitch deck, no obligation.