About

I am Antony Kanagu, a senior IT leader based in the Chicago area. I lead two teams at an audio technology company: one responsible for cloud infrastructure across AWS and Azure, and one that delivers generative AI solutions for the business. My background spans more than twenty years in software and cloud engineering, most of it on AWS.

This blog is where I write up what I learn from building agentic AI systems, in production at work and in side projects at home. The posts are field reports rather than tutorials. Most of them come from a real system I built, a benchmark I ran, or a governance problem I had to solve, and they include the numbers and the failure modes, not just the parts that worked.

The recurring themes: multi-agent architectures on Amazon Bedrock, the Strands SDK, MCP, retrieval and evaluation, cloud cost governance, and the security and compliance questions that come up when AI agents move from demo to production. One principle runs through everything I build. Deterministic code computes, and LLMs narrate and judge, never the reverse.

Some of the systems behind the posts: a competitor pricing intelligence platform, an AI cycling coach built on the Strava API, an automated security threat briefing agent, and an invoice processing pipeline with confidence-based human review.

Outside of work I am usually on a bike somewhere on the trails northwest of Chicago.

If something here is useful to your own work, or you spot an error in one of the benchmarks, I would genuinely like to hear about it. You can reach me on LinkedIn.