Data Platform
Company-wide GCP Migration
Gravel AI
Led the database migration to Google Cloud and built real-time and batch ETL pipelines with Python, Airflow and SQL.
85% lower infra cost (~£250K/yr), 40% faster processing
Data Platform
DBMS & AWS → GCP Migration
Al Jazeera Media Network
Moved to a new DBMS and migrated infrastructure from AWS to GCP, with automated Looker-based validation checks.
$500K/yr saved, 30% lower refresh latency
Experimentation
Partner Feature A/B Test
UST
Designed and ran an A/B test on new product features, alongside a ground-up Amplitude event taxonomy and tracking plan.
21% partner performance uplift, ~1.5 min faster shopping
Experimentation
Org-wide A/B Testing Framework
Gaana.com
Built a framework and tool standardising experiment design, execution and analysis; used it to improve the AutoQueue recommender.
10% growth in Listening MAU
Experimentation
Mobile Event Tracking & Validation
Al Jazeera Media Network
Designed 250+ Amplitude events with an automated tracking and validation pipeline, and defined Reforge-based growth metrics.
250+ validated events powering growth experiments
Experimentation
Retention & Checkout Experiments
Honasa Consumer (Mamaearth)
A/B tests across onboarding, checkout and CTAs, plus category-segmented retargeting experiments and attribution modelling.
90-day retention 16% → 21%, +2.3% conversion
Machine Learning
Fraud Detection Engine
Zomato
Combined a rule-based decision tree, z-tests for outliers and SVM classification; partnered with payments on anomaly flagging.
Reduced fraud and lifted revenue
Machine Learning
Lead-Scoring Model
MariaDB
Regression-based model prioritising high-potential leads, fed by Salesforce → BigQuery pipelines.
18% better sales efficiency
Data Platform
AI Data Automation with n8n
MariaDB
n8n workflows moving Gemini AI output into Google Sheets and BigQuery, plus dbt/Fivetran reporting pipelines.
10+ hrs/week saved, 60% less manual reporting
Machine Learning
Metric Anomaly Forecasting
Gaana.com
Prophet-based forecasting for early anomaly detection and a correlation-based simulation of metric interdependencies.
Earlier detection of product metric drops