WorkforceGuard AI

EU workforce intelligence and pay-transparency compliance platform for HR, people analytics, and strategy teams.

Research Paper

This system is described in:

Vamseekar, M.S. (2026). Why Tight Labour Markets Do Not Close Gender Pay Gaps: Evidence from a 20-Country Eurostat Panel.
Independent Researcher. May 2026.
SSRN: [under review] | MPRA: [under review] | Zenodo: [under review]

Live dashboard: https://workforceguard-ai.vercel.app
Keywords: gender pay gap · pay transparency · EU Pay Transparency Directive · labour market tightness · Eurostat · composite indicators

---

Live at → workforceguard-ai.vercel.app

---

Demo

Command Centre — signal scores, executive brief, live market indicators	Market Intelligence — unemployment trend, vacancy by sector, gender pay gap charts
Compare — side-by-side country and sector benchmarking	Pay Analysis — company vs market pay gap with derived scores
Govern & Export — hash-chained audit log and compliance evidence pack	AI Analyst — natural language questions with grounded evidence

---

What it does

WorkforceGuard turns public EU labour-market data and internal company payroll into decision-ready intelligence — with a full compliance audit trail built in.

• Labour market dashboard — employment rate, unemployment, job vacancies, and gender pay gap across all 27 EU member states and 13 NACE sectors, sourced directly from Eurostat
• Benchmark-aware analyst — ask natural-language questions; the copilot answers with grounded evidence, provenance citations, and benchmark context (prior period, EU average, or peer group)
• Country × sector comparison — side-by-side delta table with auto-generated narrative synthesis across any two geographies or sectors
• Pay transparency review — upload internal payroll; the platform blends it against market benchmarks and surfaces review items flagged under the EU Pay Transparency Directive
• Governance and audit log — every decision (approve, override, reverse, export) is written to a SQLite-backed hash-chained event log for legal-grade evidence packs
• Compliance evidence pack — one-click export of all metrics, provenance, and governance events as a structured bundle ready for regulatory filing

---

EU regulatory context

The EU Pay Transparency Directive (2023/970) requires employers to report and justify pay differences by May 2026. CSRD and Article 9 SFDR impose parallel workforce disclosure obligations. WorkforceGuard is designed specifically for this compliance surface — not generic HR analytics.

---

Architecture

Eurostat API  ──►  Python ingestion  ──►  DuckDB  ──►  dbt models
                                                           │
                    Internal payroll (CSV/upload) ─────────┘
                                                           │
                                                     FastAPI service
                                                           │
                                              React + Vite dashboard

Data layer: Eurostat LFS, JVS, and SES ingested as Parquet → staged and modeled in dbt (staging → core marts → internal marts → public company benchmarks) → served via DuckDB at query time, no database server required.

API layer: FastAPI with a single AnalyticsRepository that resolves filters, computes metrics, assembles evidence bundles, and writes governance events. Deployed on GCP via Docker.

Frontend: React 18 + TypeScript + Vite, TanStack Query for data fetching, Recharts for time-series visualisation, Tailwind CSS. Deployed on Vercel.

CI/CD: GitHub Actions — push to main triggers parallel deploy to GCP (backend via SSH + Docker) and Vercel (frontend).

---

Data sources

Source	Dataset	Coverage
Eurostat Labour Force Survey	Employment rate, unemployment rate, gender pay gap	EU27, 2019–2024
Eurostat Job Vacancy Statistics	Job vacancy rate by sector	EU27, quarterly
Eurostat Structure of Earnings Survey	Pay gap by sector and occupation	EU27
EGAPro (France)	Company-level gender pay index	French listed companies
UK Gender Pay Gap Service	Company GPG disclosures	UK listed companies

---

Key engineering decisions

Hash-chained governance log — audit events are chained by SHA-256 so any tampering is detectable. The chain integrity status is shown in the dashboard and verified on every API call.

Evidence provenance on every metric — every number displayed traces back to its Eurostat source ID, dataset version, formula version, and whether human review is required. This is structural, not cosmetic.

Benchmark-aware copilot — the analyst endpoint selects the appropriate benchmark basis (prior period / EU average / peer group) based on data coverage, then wraps every answer with the benchmark confidence and a coverage note. It refuses to answer confidently when coverage is partial.

dbt layered modeling — staging → core → internal → public company mart separation means the EU reference layer and company-specific layer are independently testable and replaceable.

---

Local setup

Prerequisites: Python 3.11+, Node 20+, dbt-duckdb

# Clone
git clone https://github.com/SVamseekar/workforceguardai.git
cd WorkforceGuard-AI

# Backend
pip install -r dashboard/backend/requirements.txt
cd dashboard/backend && uvicorn main:app --reload --port 8000

# Frontend (separate terminal)
cd dashboard/frontend && npm install && npm run dev
# → http://localhost:5173

To run the dbt models against local data:

cd analytics
dbt run
dbt test

---

Stack

Python FastAPI DuckDB dbt Parquet React TypeScript Vite TanStack Query Recharts Tailwind CSS Docker GCP Vercel GitHub Actions