S

shell

Recruiters deal with hundreds of unstructured resumes in PDF and Word formats. Extracting consistent, comparable data requires expensive cloud AI APIs or tedious manual effort. This project proves that a local, quantized small language model (SLM) running on a commodity CPU can do the same job — privately, cheaply, and completely offline.

LedgerLens Drop in a photo or PDF of a receipt/invoice and get clean, structured ledger data — merchant, date, tax/GST, line items, and total — as JSON, exportable to CSV. Runs entirely on CPU and works with the network switched off.