Portfolio polish: complete model evaluation, coverage, security evidence, and reproducible quickstart

[CoreyLeath-code]

Context

A senior AI engineer recommended converting the repository's remaining caveats into verified evidence and adding more project-specific engineering judgment.

Goals

• Run a complete reproducible evaluation using the intended dataset/model artifact.
• Report appropriate metrics such as per-class precision/recall/F1, macro/weighted averages, confusion matrix, and latency where supported.
• Document dataset split, preprocessing, seed, hardware/software environment, and known limitations or bias risks.
• Verify test coverage and enforce a realistic threshold in CI.
• Confirm code, dependency, secret, and container security workflows produce visible results.
• Measure Docker image size and document the image/tag and command.
• Validate the quickstart from a clean clone, including model/data acquisition or a safe demo fallback.
• Surface the strongest verified metrics near the top of the README.
• Add Engineering Decisions & Trade-offs covering class imbalance, real-time inference, false predictions, privacy, and the next production step.
• Fix typos and filename inconsistencies.

Acceptance criteria

1. Evaluation can be reproduced using documented commands.
2. The README clearly distinguishes measured metrics, demo behavior, and planned work.
3. CI reports tests and coverage and enforces the documented threshold.
4. Clean-clone quickstart succeeds without undocumented local state.
5. Security and container evidence is visible.
6. No performance, fairness, or production claims are invented.

Submit changes through a focused pull request with validation commands and results.