04 / LOGISTICS / FLEET DATA GOVERNANCE

Replacing informal card handoffs and untracked vehicle assignments with a system the team is required to keep honest.

Client: Maharashtra-based transportation firm

Tags: Logistics · Fleet Management · Data Governance · Audit Trail · FastAPI

// Context

The client runs trucks against fuel and cash sources — oil-company cards, credit at partner pumps, cash-advance cards — reconciled entirely by hand in Excel by a team of three. Underneath the reconciliation problem sat a governance problem: which card was on which vehicle, and who was driving it, drifted constantly and informally. Cards got handed off without anyone logging it, vehicles moved in and out of rental status, and there was no system requiring anyone to declare a handover when it happened — so nobody could say with confidence whether a card firing for a given vehicle was legitimate or a policy violation waiting to be discovered at month-end.

// Our Solution

A governed master-data system built specifically to stop tolerating that drift. Every vehicle, driver, card, and pump lives in a full-ceremony record — no inline shortcuts, no partial entries accepted where the business rules say a field is required. Sitting on top of that master data is an append-only assignment log: every card handover and driver reassignment has to be declared at the moment it happens, and the record can never be edited or deleted afterward.

// Technical Details

Master Data

Five governed entities — Clients, Vehicles, Drivers, Cards, Pumps — each enforced through full-ceremony forms and matching API-level validation. Ownership status has no default and must be explicit; mileage entries require a complete weight-class region; a normalizer converts legacy number-first registration formats into standard plate order.

Compliance Ledger

The Card/Driver assignment log is append-only by design — only creation and read endpoints exist, with no update or delete path, so the audit trail itself can't be altered after the fact.

Stack

FastAPI + asyncpg over Postgres, raw SQL with the schema file as the DDL source of truth, and a React/TypeScript frontend covering every master-data page plus assignment declaration and history.