May 2024 - Current
Aug 2022 to May 2024
May 2021 to Jul 2022
Sep 2019 to May 2021
Jan 2017 to Aug 2019
Staff Data Scientist (Supply Chain, ETA Prediction )
• MCMH Approach: Led development of an advanced ETA prediction system using innovative MCMH Regressor + Classification approach, enhancing delivery time accuracy by 28% across a regional distribution network serving 150+ enterprise clients
• ETA forecasting methodology: Redesigned ETA forecasting methodology by incorporating machine learning models and clustering techniques tailored to Origin-Destination (OD) patterns, achieving a 22% reduction in customer complaints and generating $450K in annual operational cost savings
• Similar Shippers Identification: Developed a custom in-house algorithm to intelligently identify and categorize shippers based on delivery patterns, geographic routes, and operational characteristics, enabling more precise logistics optimization and partner segmentation
• Data pipeline optimization: Orchestrated data pipeline optimization that reduced processing time by 35% while maintaining prediction accuracy, significantly improving system scalability for growing client base
• Monitoring: Spearheaded a cross-functional team of 8 to architect and launch unified monitoring dashboards for both shadow and production ML deployments, integrating real-time drift-detection alerts; slashed PM analysis time by 83 % (from 2 h to 20 min)
Senior Data Scientist ( Digital Advertising, User targeting )
• Real Time Campaign Selection: ML-driven customer-level scores using CatBoost and Logistic Regression for precise advertisement targeting, projecting a 12% monthly revenue boost. Instant campaign selection ensures optimal ad delivery, maximizing performance (CTR and eCPM).
• GenAI POC for revenue uplift: Implemented XGBoost model integrating visual creative & advertiser features using GPT-4 text & vision API. POC saw up to 13% performance gain. This led to significant impressions inventory saving, potentially yielding ~5% monthly revenue increase.
• Audience Targeting: Designed a model to segment and target users akin to high-performing ones in advertiser or category. Aim was to enhance campaign result and advertiser ROI. ML segments showed ~60% higher CTR than rule-based, boosting our platform’s CTR by 15%.
• Customer Persona: Identified customer personas for Paytm Ads and created cohorts of customers based on behaviour on Paytm app and ads. This helped us to understand hidden patterns, good/avg customers and active/dormant customers for 1p and MOA.
Data Scientist, Adobe.com (Search Engine Optimization, SEO)
• Migration Tracking: Analyzing daily/weekly traffic change in Adobe.com and highlighting it to leadership on regular basis. This helped to track the migration of Adobe’s subdomains and migration’s impact on conversion and revenue.
• Outreach project: Created an ML based model to understand the impact of outreach on organic traffic. This model is used by managers to estimate future efforts and ROI on their investments in backlink acquisitions. Estimated revenue jump was around $50k per month.
• Anomaly and Contribution Analysis Model: Managed a team of five including two interns to create an ensembled model using Isolation Forest and SGD to detect anomaly in Adobe’s traffic & orders based on weekly KPIs and identify contributing factors.
• Position Tracking Algorithm: Developed a simple but efficient Adobe internal algorithm using Random Forest and K-Means to understand changes in URL position over period of a time, which helped understand our SEO efforts over time.
• Identifying Similar Keywords and Creating Themes: Used Word2Vec to create cosine similarity-based model to create themes out of 1M keywords. These themes were used to identify new SEO opportunities.
• Multiple Analytics Projects: Handled multiple insights and analytics projects which were shared with higher leadership to understand the impact of various efforts in SEO to bring more organic traffic. These efforts helped forecast a revenue of $4M for CC & DC Adobe clouds.
Senior Data Scientist, UK Based Client (Ecommerce + Retail)
• Point of Market Entry: Built a module to identify clearly defined entry point for customers. The target was to understand which products customer prefer when they are buying into specific category.
• Front of Store Missions/Themes: FOS includes multiple categories making it hard to analyze, needs of customers. Using NMF and Clustering on 7M baskets, we created a mission view for vital FOS objectives, revealing shopper engagement patterns and basket composition.
• POS Customer Segmentation: Built a model to tag customers based on available historic customer segmentation that splits customers based on life stage & affluence. This helped to understand profiles of POS Customers
• Identifying Cannibalization: Client reported inverse sales behaviour after they launched new products. We hypothesized this problem as cannibalization and justified it using Natural Language Processing model based on string matching algorithms & user behaviour analysis.
Data Science and Analytics (Retail + Healthcare)
• Apple Sales Forecasting: Worked on ML modelling for Apple Sales to forecast apple devices sales for upcoming months.
• MySupply project: Tracked routine changes in Apple devices for every device (outside India Region). The task was to coordinate with business and send them regular insights about the impact of those changes in business.
• Change Ownership: Led changes for client software requirements, managed a diverse team and ensured smooth business processes.