Skip to content
View amira-ghazy's full-sized avatar

Block or report amira-ghazy

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
amira-ghazy/README.md

Amira Ghazy

Measurement scientist who builds. I work where human judgment meets automated scoring — model evaluation, people analytics, and assessment systems that have to be defensible, not just accurate.

My background is a master's in industrial-organizational psychology with an advanced quantitative methods focus, plus hands-on engineering with machine learning and large language models. The short version: I can train the model and tell you whether its scores actually mean anything.

Open to roles in AI evaluation, people analytics, and psychometric data science.

Selected projects

Does the Model Rate Like a Human? — an LLM scorer for open-ended responses, validated like an assessment: inter-rater reliability (ICC, weighted kappa, Krippendorff's alpha) and a fairness audit (Cohen's d, four-fifths adverse-impact rule). LLM eval · psychometrics · fairness

Predicting Attrition Without the Black Box — turnover prediction built to be explained and audited: logistic odds ratios tied to turnover theory, permutation importance, and per-group error-rate parity. applied ML · interpretation · fairness

What Employees Actually Say — open-ended survey comments turned into themes, then linked to engagement so the output is what to act on, not a word cloud. NLP · topic modeling · constructs

Toolkit

Measurement reliability & validity · factor analysis · IRT · adverse-impact analysis ML scikit-learn · model evaluation & interpretation · experiment design LLM/AI Anthropic & OpenAI APIs · LLM-as-evaluator pipelines · rubric design · RAG Eng Python (pandas, NumPy) · Git · reproducible analysis · R

✉️ miraghazy1989@gmail.com · Portfolio

Pinned Loading

  1. amira-ghazy.github.io amira-ghazy.github.io Public

    HTML

  2. attrition-fairness attrition-fairness Public

    Employee turnover prediction built to be explained and audited — odds ratios tied to turnover theory, plus per-group error-rate parity.

    Python

  3. employee-comments-nlp employee-comments-nlp Public

    Turn open-ended survey comments into themes, then link each theme to engagement — what to act on, not a word cloud.

    Python

  4. llm-evaluator-psychometric llm-evaluator-psychometric Public

    An LLM scorer for open-ended responses, validated like an assessment — inter-rater reliability plus a fairness and adverse-impact audit. Runs offline in mock mode.

  5. auditing-the-llm-judge auditing-the-llm-judge Public

    A psychometric audit of an LLM-as-judge: test-retest reliability, verbosity and position bias, and group invariance.

    Python

  6. evalaudit evalaudit Public

    Psychometric audits for evaluation data — reliability, agreement, bias, and measurement invariance in one call. A drop-in audit for LLM-judge and human-rating pipelines.

    Python