Field notes from building production AI systems — architecture decisions, failure modes, and lessons from shipping.
Real-time voice AI talk show with <100ms latency. Multi-agent orchestration, audio pipeline design, and state management challenges.
What I learned building a multi-tenant AI platform that handles real workloads—from architecture decisions to the problems nobody warns you about.
Addressing a common concern from technical leaders: is AI-assisted development producing lower quality code? Here's what I've learned from shipping production systems.
Building an AI-powered legal research platform for the Greek legal system—the unique challenges of domain-specific AI and why RAG alone isn't enough.
The demo worked perfectly. Six months later, you're still 'almost ready' for production. Here's why AI projects get stuck and how to actually ship.
How I built a live broadcast system where AI agents debate topics in real-time—the architecture, the challenges, and why it's the ultimate AI engineering flex.
Retrieval-Augmented Generation sounds simple: embed documents, retrieve relevant chunks, generate answers. So why do most RAG systems disappoint in production?
As someone who uses AI daily to write code, here's my honest take on what AI can and can't replace—and what it means for hiring technical talent.
Everyone wants a voice AI agent. Few understand what's actually possible today. Here's a grounded assessment of voice agent capabilities and limitations.