Hardware AI Lab

This repository is my public build log for learning and applying AI as a hardware engineer. Each project is designed to prove a complete engineering loop:

1. Define a real physical problem.
2. Instrument it and collect data.
3. Build a baseline before adding AI.
4. Measure accuracy, latency, power, cost, and failure cases.
5. Publish the design decisions, code, test results, and demo.

Current Build

Edge Vibration Anomaly Monitor

Detect unusual behavior in a small motor or fan using an accelerometer and an ESP32-S3. The first software prototype extracts useful vibration features from normal and fault-like signals. Later milestones move data collection and inference onto real hardware.

• Project brief
• Latest learning log
• Run the first demo: make demo
• Run the tests: make test

Project Ladder

Stage	Project	What it proves
1	Edge vibration anomaly monitor	Sensors, DSP, TinyML, validation
2	Battery health and load-test station	Power electronics, test automation, prediction
3	PCB visual inspection assistant	Computer vision, fixtures, defect datasets
4	AI-assisted bench instrument	SCPI automation, waveform analysis, human-in-the-loop AI
5	FPGA streaming signal classifier	RTL, fixed-point design, latency and resource tradeoffs

See PROJECTS.md for the detailed project menu and selection criteria.

Portfolio Standard

Every finished project should include:

• A 60-90 second demo video or GIF
• A block diagram and a readable schematic
• A bill of materials with total cost
• Reproducible firmware, software, and setup instructions
• Test data and quantitative results
• A short section called "What failed and what I changed"
• Issues and milestones showing how the work progressed

Build Rhythm

• Commit small, understandable changes several times per week.
• Open an issue before each meaningful milestone.
• Write a weekly learning log, including failures and measurements.
• Tag working demos as releases.
• Keep unfinished work visible, but label it honestly.

The goal is not a wall of repositories. The goal is three finished projects that make engineering judgment visible.