Portfolio evidence map

This repository is designed to make engineering and research evidence easy to verify.

Capability	Evidence
Python package design	src/crisis_signal/ and pyproject.toml
Data-quality thinking	data.py, dataset card and duplicate-aware split
ML baselines	models/classical.py, training.py
Deep-learning breadth	optional BiLSTM and transformer adapters
Trustworthy AI	calibration, abstention and weighted error cost
API engineering	api.py, schemas and API tests
MLOps	DVC pipeline, MLflow adapter, Docker and CI
Monitoring	PSI and token-distribution drift
Multimodal foundations	image hashing and duplicate detection
Research integrity	benchmark evidence status and reproducibility protocol
Teaching potential	book outline, documentation and modular experiments

A recruiter or academic reviewer should be able to distinguish implemented features, historical evidence and planned research without guessing.