Subhadeep Halder

Ideated an everybody wins A/B campaign based on user demographics and win streak patterns thus improving user conversion by 9% and average user turnaround post depositing by 6%

Conducted A/B testing to compare static vs personalized cash-load tile recommendations, driving a 20% increase in deposit conversion for the personalized group

Developed comprehensiive user and campaign specific rewards dashboard, leading to seamless tracking for coin disbursals and driving 8% increase in user engagement in flight bookings

Built a merchant recommendation system using cohort analysis on location and forex transaction data boosting merchant engagement by 22% across 2L+ users in 60+ countries via hybrid filtering

Built a trip identification logic using international transaction data, enabling country-wise and seasonal spend pattern analysis for over 2L customers, leading to a 10% improvement in marketing efficiency

Led an A/B test to evaluate the impact of removing the passport verification step, resulting in a 13% increase in conversion rate by reducing user friction in the bank onboarding funnel

Devised accurate battery replacement forecasts for each year up to 5 years using battery SoH, charging and mileage data, achieving 87% predictive accuracy in the first year's assessment

Automated data ingestion pipelines for e-scooter performance testing using SQL, Azure Data Factory, and Databricks, enabling efficient data processing and transformation before ingestion into Power BI dashboards, reducing manual effort by 30% and improving analytics speed by 40%

Designed and implemented an AppScript solution automating bike rentals to testers, resulting in a weekly labor reduction of 5 hours

Optimized the process of uploading audio files to AWS S3 from a Node server by making it 3 times faster

Used RAKE algorithm for keyword-extraction from the texts transcribed from audio files

Implemented HuggingFace Transformer for emotion recognition on texts and compared with HMM model, calculated the polarity and subjectivity of text files using TextBlob and plotted the results using D3.js