Looking for a job

Som Subhra Ghosh

@somsubhraghosh

Message

I build large-scale graph recommendation systems and production machine learning platforms.

Japan

Message

At PayPay, I build graph-based recommendation systems for coupons, post-payment offers, and merchants. I use FastRP and Personalized PageRank to rank relevant offers in real time, including a merchant recommender spanning 35+ million users, nearly 2 million merchants, and 2 billion monthly transactions.

I've also delivered optimization and ML products across e-commerce and fintech, including banner models that lifted CTR by 17% and conversion rate by 1.2%. My work spans distributed GPU pipelines, ranking, feature engineering, API serving, monitoring, graph neural networks, and operations research.

Experience

Work history, roles, and key accomplishments

Current

Senior Data Science Engineer

Current

PayPay

Sep 2023 - Present (3 years)

Developed graph-based recommendation systems for campaign reranking, post-payment offer discovery, and merchant recommendations using FastRP, Personalized PageRank, and cuGraph on multi-GPU clusters.

Python SQL Spark AWS Graph ML C C++FastRP PageRank CuGraph Dask fastAPI Optuna

Research Scientist

PayPay

Dec 2022 - Aug 2023 (8 months)

Improved coupon relevance using wide-and-deep neural networks and optimized multi-page engagement for an MNO website through clickstream analysis and engagement modeling.

Wide and Deep Neural Networks Clickstream analysis Session Behavior Analysis Digital Engagement Models

Staff Data Scientist

Rakuten

Aug 2020 - Dec 2022 (2 years 4 months)

Built stochastic bandit models for banner optimization achieving 17% CTR lift and 1.2% conversion-rate lift, and trained GCNs for graph partitioning of GPS-location data.

Stochastic Bandits Contextual Bandits Offline RL GCN Graph Partitioning GPS Data Analysis

Data Scientist II

Amazon

Apr 2019 - Aug 2020 (1 year 4 months)

Developed browse node propensity models and event propensity models to predict purchase probabilities and optimize marketing targeting, and quantified the impact of Google Shopping Ads on customer engagement.

Propensity Modeling Machine Learning Customer Engagement A B Testing Causal Inference

Senior Statistical Analyst

WalmartLabs

Oct 2017 - Apr 2019 (1 year 6 months)

Implemented customer probabilistic matching and propensity models using Azure ML, and applied LIME for explainable AI to improve model-quality review.

Azure ML Boosted Trees Feature Engineering API Deployment LIME Explainable AI

Senior Business Analyst

WalmartLabs

Jan 2016 - Oct 2017 (1 year 9 months)

Optimized vehicle routes for large appliances using ant-colony optimization, built a TensorFlow/Keras service for address intelligence, and created dynamic beat optimization with convex optimization.

Ant Colony Optimization TensorFlow Keras Convex Optimization PuLP CPLEX Route Optimization