Skip to content
All work

04

INTERNAL PORTAL & REPORTING AUTOMATION 

Rebuilt how 500+ people at ONGC Videsh found documents and received their weekly numbers, and took the manual assembly of those numbers out of the loop entirely.

Role
Engineering Intern, ONGC Videsh
Stack
Python · Pandas · openpyxl
Year
2025
Status
Shipped
01

Context

An intranet portal that more than 500 people depended on to find documents and circulars, and which had grown by accretion until finding anything meant knowing in advance where it had been filed. People asked colleagues instead of searching, which is the clearest possible signal that a tool has stopped working.

Running alongside it was a recurring reporting cycle assembled by hand every week: pull the same extracts, paste them into the same workbook, reconcile them against last week, format, send. It cost a full working day a week and the errors it produced were the quiet kind, a mistyped figure that nobody catches until a decision has already been made on it.

Both problems had the same root. Work that was perfectly well understood had never been written down as a process, so it lived in whoever happened to be doing it that week.

02

Approach

  1. 01

    Watched the actual weekly routine end to end before changing anything, because the version people describe and the version they perform are rarely the same document.

  2. 02

    Rewrote the assembly step in Python with Pandas and openpyxl so the extracts are read, joined and written into the finished workbook without a human copying a single cell.

  3. 03

    Moved the transformations that genuinely belonged in Excel into Power Query, so the people who own the report can still open it, see what it does and change it without asking an engineer.

  4. 04

    Added reconciliation against the previous period so a figure that moves unexpectedly is flagged while the report is being built, not after it has been read.

  5. 05

    Reorganised the portal around how people actually look for documents rather than the department that filed them, and made search reach the content instead of only the titles.

  6. 06

    Wrote the runbook alongside the code, since an internship ends and an automation nobody can rerun is just a slower version of the original problem.

03

Results

people using it every day
500+
hours of manual work saved weekly
10+
typing errors after rollout
0
command to rebuild the whole report
1
04

Stack

  • Python
  • Pandas
  • openpyxl
  • Power Query
  • Advanced Excel
  • Internal Tools
Diagram soon ✱

Architecture sketch: INTERNAL PORTAL & REPORTING AUTOMATION

Previous project

COLLEGE PREDICTOR DATA VALIDATION

Next project

DATA DESK