CASE/04 / Manufacturing
IndiaFour decades of HR paper, made searchable
1,527 active and 178 inactive records structured for search and export.
01 / THE ASK
The decision behind the brief.
Make four decades of head office and central staff records usable for a move to a modern HRMS.
02 / WHAT WE FOUND
What the work revealed.
A listed Indian enterprise with more than 10,000 employees held the records in hardcopy. Changing formats, languages and handwriting meant a single extraction template could not describe the archive. Duplicates added another reconciliation problem.
03 / WHAT WE BUILT
A paper archive becomes usable.
Route before extraction
Documents are captured and classified. Vision-language models help route them to the appropriate extraction engine and schema, followed by validation.
Find the record
Structured metadata supports search, filters and CSV export. A compliance checklist helps staff inspect the required records.
Ask in the working language
An assistant supports Hindi, English and Hinglish queries against the digitised records.
Placeholder for an approved, masked view of the delivered work.
04 / THE RESULT
1,527 active and 178 inactive records structured for search and export.
The live system contains 1,527 active and 178 inactive records structured for search and export. These are the delivered record counts, not the enterprise’s total workforce.
[CONFIRM: delivery year, current record-count date and remaining application stack]
05 / ENGINEERING DETAILUnder the hood
- Scans and document classification
- Routed extraction and validation
- Search, export and assistant
The architectural step is classification and routing to the appropriate extraction engine and schema. [CONFIRM: routing criteria, confidence thresholds, human review controls and storage]
Workflow sketch. Unconfirmed components are marked explicitly.
- OCR
- Vision-language models
- [CONFIRM: remaining stack]
A GOOD PLACE TO START