Founder
Mahmoud Zalt: solo engineer behind Sistava. AI builder, software engineer, backend specialist. Founder of Sista AI (registered in the Netherlands, KvK 90724933). Trust is built on track records, not promises.
It started with a question
In my first year of university I came across the Turing Test: the idea that a machine might converse so naturally a person could mistake it for human. The question stayed with me. I was already drawn to building things, and the prospect of building one of those machines pulled hard. The skills were not there yet, so I parked the idea and went back to the systems I knew how to build. It sat in the background for years, waiting for the tools to catch up.
I have always been a builder
I have been writing software professionally since 2010, specializing in backend, infrastructure, and developer tools. Over the years I have built multiple open-source projects that achieved massive global adoption. Most notably, Laradock has crossed 2.5 million downloads worldwide, and Apiato has become the go-to framework for agencies building modular, scalable APIs. By my second university year I had built and sold my first real product, and that build-and-ship loop has shaped my work ever since.
Then ChatGPT arrived
In late 2022, ChatGPT shipped and the old Turing-Test question came back sharper than before. I committed to AI full time. Not once, but every time the next layer of the problem opened up I followed it further in. Each layer revealed the one beneath it, and the work kept getting more interesting.
First try: meeting people where they already talk
My first instinct was to put AI in front of non-developers, in the apps people were already in: WhatsApp, Telegram, day-to-day chat. I built a way for an assistant to live there and answer real questions in plain language. The product was small, but the lessons were not. It taught me how these models actually behave in production: where they shine, where they hallucinate quietly, and how much of the experience is the framing around them rather than the model itself.
Second try: making the web itself talk back
Next I wanted any website to become talkable: open a page, ask it something, have it act on the page for you. I built a browser extension and started handing the AI more of the whole browser to operate. The more control I gave it, the more I saw the real frontier was not single-shot answers. It was longer, multi-step work that crossed pages, tabs, and tools. The goal moved from smarter pages to actual agents, and I followed it there.
The wall I had to hit
The framework I was building on was never designed for long-running autonomous work. I tried to stretch it into something that could run real tasks for real minutes, and it kept dead-ending. After roughly eighteen months pushing on those limits I accepted the foundation itself had to change. A painful stretch, but the most valuable education I have had as an engineer.
Could AI just do my own job?
The honest version of what I was chasing was personal. I was already leaning on AI for half my day: drafting emails, summarizing what I did not want to read, ghostwriting first passes. What I wanted was an AI employee I could brief like a normal person, connect to real tools, and hand actual work to. Not a demo, not a chatbot inside a workflow. Something that would hold up once the work got messy and the answer mattered.
Built so AI can maintain it
Every component in Sistava is pluggable: I can replace, extend, or rewrite any piece without unraveling the rest. I designed it that way on purpose, because I refused to inherit the limits of my last framework, and because the platform itself has to be something AI can help maintain over time. The architecture is the product as much as the features are.
Three and a half years, four products
Sistava did not arrive in a single sprint. It took three and a half years and four different AI products to get here, and none of them were wasted. Each one taught me a piece of the puzzle: prompting, memory, workflows, browser control, voice, reliability, and the speed at which AI falls apart when execution becomes real.
I hire my own AI too
The product I ship to you is the product I run my own company on. My own AI employees handle the programming work through a custom coding harness I built, and others handle finance, legal, and marketing. Sistava is the engine that lets a solo founder cover ground a single human cannot.
And then this one worked
Sistava is live, used by real people, and improved every week based on what those users actually run into. It is bootstrapped and still solo. That keeps it honest. Every decision has to earn its place, every feature has to survive contact with real work, and there is no investor narrative bending the product.
What I actually want
I am not chasing an exit or an IPO. The part of this I genuinely love is the engineering: getting a hard system to behave the way it should. My hope with Sistava is simple: that it earns enough to let me keep working on it, for as long as the problem stays interesting. That is the entire ambition.