Who we are
Building operational infrastructure for AI agents
ZIZKA AI S.L. is an EU-based company in Málaga, Spain. We build systems that help teams ship AI agents with replay, causal lineage, and drift detection — not just scattered logs.
Our flagship product: ZizkaDB
ZizkaDB stores every agent step as a linked event. When something breaks in production, you need to know why an agent chose an action — not guess from partial traces. ZizkaDB gives you session replay, decision chains with why(), point-in-time rewind with at(), and alerts when behavior drifts after a prompt or model change.
The same product runs on managed cloud at db.zizka.ai or self-hosted open source under AGPL-3.0. Python, TypeScript, MCP, and REST — your stack, your choice of hosting.
Mir Arshad Talpur
Founder, ZIZKA AI S.L.
I am a founder with multiple startups and a background at the intersection of software engineering, artificial intelligence, and systems thinking. I am an NVIDIA-certified Agentic AI professional with years of hands-on experience delivering complex software across the US, EU, and Gulf markets.
ZizkaDB comes from a simple observation: teams shipping agents to production need operational data — replay, lineage, and drift — not another thin wrapper on a language model. We chose the harder path: a purpose-built store for agent behavior that is predictable, auditable, and aligned with how engineers actually debug production systems.
Saad Amjad
Founding Engineer
Full-stack engineer with seven years shipping mobile and web products to millions of users. At Washmen — the UAE’s award-winning laundry platform with 1M+ active users — I owned features end-to-end across React Native apps, eight partner-branded PWAs, the Sails.js backend, and AWS infrastructure.
Earlier I was a core engineer at Retailo (B2B marketplace, KSA/UAE) and built Hao (wellness app, Riyadh) solo from MVP to App Store launch. At Zizka I lead product engineering across Next.js, Node, PostgreSQL, and cloud — with the same rigor I applied shipping to production at scale.
Dr Franz Scholder
Strategic Scientific Advisor
PhD in Mathematics (Università degli Studi di Milano-Bicocca), with published work in algebraic topology, topological fluid dynamics, and geometric flocking (IEEE Transactions on Automatic Control, Journal of Singularities). Former Head of Global Underwriting at Dennemeyer IP, where he built a quantitative underwriting function from scratch and delivered €2M+ in additional annual profit.
Now an independent AI consultant in Trier, Germany, he designs and ships LLM systems — voice agents, RAG recommenders, and research tools validated with real users. Advisor to ZIZKA AI S.L. (ZizkaDB) from July 2026, bringing mathematical rigor and production AI experience to how we reason about agent behavior at scale.
Why the name Zizka
The company is named after Jan Žižka, a historical general who used strategy and technology — the war wagon — to give a small, resource-constrained force leverage against much larger armies.
That is the bet behind ZizkaDB: small teams and solo builders should have the same operational visibility as large orgs, through better infrastructure rather than bigger budgets.
How we think about agent systems
Production agents need guardrails, human oversight, and clear audit trails. We align with responsible AI frameworks — including the EU AI Act direction — and build tools that make agent behavior inspectable rather than opaque.
ZizkaDB is not about hype or full autonomy for its own sake. It is about giving engineers the data they need to trust, debug, and improve agents over time.
Learn more about the company on zizka.ai, or visit the ZizkaDB product site for technical docs.