Kevin Houston
Writing

Notes on building in practice.

First-hand implementation notes on applied AI, enterprise APIs, customer-data architecture, and production delivery.

Delivering a Bounded AI System: Discovery, Deployment and Recovery

A first-hand client case: customer-data discovery, bounded model roles, evaluation, production recovery, and bespoke client training.

Read case study

Solutions Engineer vs. AI Deployment Engineer: What Changes?

The roles share discovery, integration, and delivery responsibilities. AI adds decisions about variable outputs, permissions, evaluation, and ongoing ownership.

Read more

What a Production AI Agent Deployment Actually Requires

Five implementation decisions: a bounded workflow, approved sources, enforced action boundaries, behavioral checks, and an owner after launch.

Read more

Building Source, a Local-First Transcriber for Raw Transcript Ownership

I did not build a new transcription model. I built a local-first wrapper around faster-whisper because the existing meeting tools optimize for summaries, and I wanted the raw transcript source file. The writeup covers macOS packaging, mic and system audio capture, model tradeoffs, and what I learned about input quality versus model size.

Read more

Project case studies

Implementation decisions, ownership, and evidence from my applied AI projects.