Senior Backend Engineer · Java and Go · AI engineering
I build and run backend services where privacy, correctness and uptime matter: banking channels, payments, and today a device network at very large scale. On the side I build with LLMs, MCP and agents, and I turn what I learn into interactive guides anyone can use.
Malmö, Sweden55.60°N 13.00°EWriting backend code since 2006
/lab
Interactive guides in three tracks. Each one runs the real logic step by step, often next to the Java code, so you can see why it works. Made for new team members, mentees, and anyone who wants a refresher.
/about
I've been building backend systems since 2006. My first fifteen years were in Iran: internet and mobile banking channels used by several banks, a cryptocurrency exchange, and a fintech platform where I led the move from a PHP monolith to Java and Go services handling payments, credit and top ups.
Since 2021 I've worked in Sweden. At Klarna in Stockholm I helped design and build an automated Right to be Forgotten platform and worked on a decision engine used across the company. Today I'm in Malmö, working on the backend behind Apple's Find My.
What ties it together is systems where correctness, privacy and availability are not optional. In my own time I work on the newer half of modern backends, putting LLMs to work inside real services with MCP, RAG, agents and proper evaluation.
/experience
Since 2021 I've been a permanent engineer at 0+X; my Klarna and Apple roles are through 0+X, where I also mentor in the Compass program and share AI-assisted development practices. B.Sc. in Software Engineering, Sheikh Bahaei University, Isfahan, 2006.
/services
Twenty years of building and running backends, available to your team in focused pieces of work: a review, an audit, a workshop, or a few hours of advice a month. Tell me about your system and what you want to change, and I'll tell you honestly whether I'm the right person for it.
/now
Kairos reads unstructured financial news and turns it into scored, structured signals. It combines LLMs with structured outputs, retrieval over a vector store, tool-using agents connected through MCP, and an evaluation harness that keeps the scores honest. Built in Java and Python.
A research project, not investment advice. Happy to talk about the architecture.
/toolbox
/writing
/talks
/contact