A 60-person transcription team. Two languages. A Q1 in freefall — quality dropping, SLA targets missed, leadership convinced it was a headcount problem.
It wasn't.
- Python (Pandas/NumPy): The Engine. Used to engineer a 21,900-row synthetic dataset that modeled real-world operational volatility.
- SQL: The Auditor. Used to perform diagnostic queries and isolate the exact "Hindi Orthographic" failure points.
- Excel: The Interface. Designed a dynamic, executive-facing dashboard to visualize the strategic turnaround.
Instead of manual data entry, I built a Simulation Engine.
- Programmed "Crisis Logic" into Q1 to mirror real-world team struggles.
- Used Binomial Distribution to simulate realistic random quality audits.
- Automated the cleaning process to output a "Dashboard-Ready" dataset.
I acted as a "Data Detective" to find out why we were failing.
- Linguistic Bottleneck: My SQL queries isolated a 9.2% spike specifically in Hindi Orthography.
- SLA Validation: Built logic to calculate attainment % at the worker level, identifying under-resourced team segments.
I translated technical rows into a Decision-Support Interface.
- Normalized Metrics: Converted millions of units into a clean "SLA Attainment %" for easier stakeholder digestion.
- Visual Strategy: Used Radar Charts for risk mapping and Combo Charts to show the moment we "broke through" the 100% target in Q3.
- 22% Quality Surge: Following a data-backed resource reallocation.
- 104% SLA Attainment: Moved from a "Failure State" (91%) to exceeding targets (104%).
- ZERO Headcount Increase: Solved the bottleneck by optimizing existing talent rather than hiring.
/SQL_Source: Diagnostic audit scripts./Python_ETL: Data generation and cleaning logic./Dashboard: The final.xlsxExecutive Performance Summary./Dashboard_Preview: The dashboard screenshot.
Based on real operational work at Sigma.AI — reconstructed with synthetic data for portfolio purposes. Figures are accurate. Identifiers removed.
Mekhala Saxena · LinkedIn · saxena.mekhala97@gmail.com