04 — Workforce & Payroll
Timesheet Processing & Billing Audit
Timesheets arrive as spreadsheets, PDFs, photographs and email bodies. All of it becomes billable hours.
A timesheet automation and audit system for a staffing workflow. Files arrive through upload or a monitored mailbox in every format people actually send — clean spreadsheets, scanned paper, phone photographs, hours typed into an email body.
The system reads them, matches employees and clients against master records, applies overtime rules, and flags anything that does not reconcile before it reaches billing.
What I built
- 01Multi-format ingestion: spreadsheets, PDFs, images and email bodies
- 02Deterministic parsers for structured files, vision models for everything else
- 03Fuzzy employee and client matching against master records
- 04Overtime rule application and billing reconciliation
- 05Background job pipeline with a queue, retries and a cost circuit breaker
- 06Review and approval interface with live processing status
Outcomes
- —Extraction service decoupled into its own deployable with no loss of throughput
- —Historical corpus replayed to find defects that had never been reported
- —An entire cache dependency removed from the architecture