AI Embedded Systems
Engineering notes

Original work, operating details, and the failure paths.

These notes explain what we built, what we tested, what remains uncertain, and what a technical buyer should verify before committing to a larger system.

Published

Start with work that exists today.

Research queue

Planned only after there is original evidence to publish.

These are editorial commitments, not empty keyword pages.

Private AI pilot success criteria

Workflow selection, data boundaries, citations, evaluation, user testing, security, and stop conditions.

Local AI hardware sizing

A measured comparison of model, quantization, VRAM, prompt size, time to first token, throughput, memory, and failure conditions.

ROS 2 and edge AI audit checklist

Topics, rates, DDS, sensor timestamps, compute budget, model latency, thermal behavior, logging, and recovery.

Private document security

Provider boundaries, local deployment tradeoffs, retrieval, retention, review, and operational ownership.