I build production AI agents and the backends they run on.
I'm Andi, a backend developer with more than 14 years of experience. I started with PHP and Laravel and I still work with them, now with a focus on AI: bringing LLMs into real systems and taking them to production.
That means agents that call a product's own APIs, retrieval that answers real questions, MCP servers that open a system to outside agents, and the evals and guardrails that show whether any of it is working.
The backend side hasn't gone anywhere: APIs built to stay fast under load, with PHP and Laravel on Swoole, and databases like MySQL, MongoDB and Redis behind them.
LLM tool calling on real product APIs, in multi-step workflows built to fail safely.
Ingestion, chunking, embeddings and retrieval, tracked end to end for grounded answers and latency.
SaaS APIs exposed to external agents, with permission-aware access.
PHP and Laravel on Swoole, with cache and query tuning to keep latency low under load.
Orchestrated workflows with fallbacks and monitoring, tuned to cut spend and response time.
Prompt and context engineering plus systematic evaluation, so answers stay grounded on sensitive data.