Public‑Service Data & AI

I work in the federal public service as a hands‑on data & AI practitioner. My day‑to‑day spans data engineering and management, applied analytics, and ML/NLP across survey and operations data. I design and review instruments and methods, build PII‑safe data pipelines with quality checks, and deliver decision‑ready dashboards and briefs; I also provide internal consulting and methodology reviews across branches.

Work shown here is anonymised and described at a thematic level to respect confidentiality and privacy‑first processing.

Key Project — AI Language‑Learning Hub (sole developer & maintainer)

  • Led the alpha‑stage build of a Python/Flask prototype for an AI‑powered public‑service language‑learning hub, translating stakeholder vision into a working app and preparing it for the 2025 Public Service Data/AI Challenge.
  • Team won the semi‑final round in June 2025.

Responsibilities (selected, categorised)

Data engineering & quality

  • Build and maintain Python/SQL ETL from extraction and cleaning to standardisation and audit snapshots.
  • Define YAML data contracts (e.g., columns.yml, config.yml) to encode row grain (one row per employee‑month), join logic, de‑identification, and data‑quality checks.
  • Contribute to schema migrations by scripting table/data comparisons and drafting test plans to validate outcomes.

Survey methodology & measurement

  • Design/review questionnaires, sampling, and measurement; align scales and weighting; document reliability/validity and reusable methodology packages.

NLP & machine learning

  • Run text analytics (word clouds, sentiment, topic models) with VADER/TextBlob/LDA/BERTopic.
  • Fine‑tune Transformer models (BERT/RoBERTa) for sentiment and topic discovery; implement in TensorFlow/PyTorch with explainability and evaluation notes.

Visualisation & reporting

  • Build dashboards and reports in Power BI and Python to turn data into decision‑ready insights; structure executive summaries and KPI frameworks.

Internal consulting & collaboration

  • Provide cross‑department methodology reviews, instrument/intake advisory, lightweight data models, and facilitation (working sessions, playbooks, guidance notes).

Privacy & governance

  • Apply privacy‑first processing and PII‑safe de‑identification; version and log QA/audit steps to support reproducibility and compliance.

Project names and proprietary terms are intentionally omitted; details are available on request where appropriate.