Building a Markdown-First Citation Pipeline for an AI Health Agent
How I replaced inline URL citations with a token-based reference system that separates machine semantics from user-visible text — and why the old approach wa...
How I replaced inline URL citations with a token-based reference system that separates machine semantics from user-visible text — and why the old approach wa...
Three ECS services from one Docker image, the env var crash that blocks every first Celery deploy, Redis configuration across environments, and health checks...
Building a 30-day backfill system with anchor-based incremental sync, chunked uploads, UUID deduplication, and the ‘device locked’ problem that breaks everyt...
You can’t process Apple HealthKit data inline. Here’s why we needed Celery, why SQS was the wrong broker for us, and the architecture that emerged.
After 14 years of building software, I’m finally writing about it.
A visual deep-dive into running Celery workers as separate ECS containers — broker topology, task lifecycle, concurrency models, and the Beat scheduler gotcha.
I have 268 markdown files in my Obsidian vault – internal docs, architecture decisions, API specs, meeting notes. Searching them with Obsidian’s built-in sea...
Part 4 of building the AI assistant. This covers the most critical issue in health AI: preventing the model from making up data that doesn’t exist.
Part 3 of the series on building a women’s health AI assistant. This covers the system prompts, safety guardrails, and behavior control that make an AI agent...
This is the first in a series documenting how I built a women’s health AI assistant using AWS Strands SDK, Bedrock, and Django. The system handles personal h...