A bit about me
I was born and raised in Shanghai, lived in London for six years, and now live and work in Stamford, Connecticut. I am a data scientist specializing in segmentation, predictive modeling, and advanced consumer analytics. With 8+ years of experience across UK and US markets, I focus on transforming complex datasets into scalable insights, machine-learning solutions, and measurable commercial outcomes.
Manager, Audience & Segmentation – Philip Morris International (PMI), USA
I bring deep hands-on data science expertise together with strategic leadership in segmentation, predictive modeling, and analytics frameworks used across multiple PMI markets. In 2024, I relocated from the global function to the US market to help shape and scale advanced analytics capabilities for the US commercial organization, supporting major consumer-facing channels (CRM, paid media, customer care, retail, and web).
- Age: 30
- Website: yurunsang.github.io
- Phone: +1-2037081241
- Now live: Stamford, Connecticut, USA
- Gender: Female
- Experience: 8+ years
- Email: yurunsang@icloud.com
- From: Shanghai, China
Beyond analytics, I’m energised by creativity and exploration - discovering new restaurants, travelling to new cities, or diving into story-driven games. At home, my Maltese Luna makes sure life never gets too quiet.
Skills
Expertise in advanced analytics, segmentation, predictive modeling, and machine learning. Strong foundation in Python, SQL, and R, with experience building scalable solutions that power commercial decision-making.
Advanced Analytics & Data Science
Segmentation & clustering, predictive modeling, ML pipelines, feature engineering, and behavioural analysis across consumer lifecycle stages.
Channel Strategy & Activation
Audience design and activation across CRM, paid media, web, retail and customer care. Prior experience building client-facing dashboards (Power BI, Tableau, Qlik, Streamlit) to translate data into insight.
Data Engineering, CDP & MLOps
Feature stores, rolling-month & time-travel logic, ETL pipelines and production ML workflows across Treasure Data, AWS, and Bitbucket/Git.
Resume
If you’d like a more comprehensive view of my experience and project history, you can download my full CV below.
Download full CV (PDF)Professional Experience
Manager, Audience & Segmentation – US Market
Mar 2024 – Present
Philip Morris International – Stamford, CT, United States
- Lead audience segmentation, CDP strategy and data activation to support the launch and growth of IQOS in the US market.
- Own the consumer data platform in Treasure Data, integrating CRM, eCommerce, paid media, customer care, retail and F2F channels.
- Design predictive and rules-based audiences including life-stage segmentation, lead-scoring propensity models, and zipcode-based “Treasure Map” segmentation for precise targeting and indirect retailer expansion.
- Develop ZYN life-stage segmentation and competitive performance analytics across US geographies.
Manager, Segmentation & Modelling – Global Function
Dec 2022 – Mar 2024
Philip Morris International – London, United Kingdom
- Led segmentation and predictive modeling for the global consumer experience function, supporting 30+ markets.
- Designed a global ML feature store with 500+ reusable consumer 360 features using rolling-month and time-travel logic in Treasure Data.
- Built predictive models including referral, churn, and flavour advisory using Python and PySpark on AWS.
- Established an MLOps framework using Bitbucket/Git to standardize workflows, governance and reproducibility.
- Enabled omni-channel activation of model outputs, driving measurable uplifts in acquisition and retention KPIs; presented the scalable framework at the 2024 Treasure Data Conference.
Early Career
Data Scientist
2021 Oct – 2022 Dec
RBS International, NatWest Group – London, UK
- Developed queueing simulation tools to support branch network optimisation.
- Built clustering models to detect business transactions within personal accounts.
- Developed mortgage churn prediction models supporting customer retention strategies.
Data Scientist
2018 Sep – 2021 Sep
CACI – London, UK
- Led GPS analytics modeling using DBSCAN and mobility clustering to derive home/work locations for thousands of users, supporting major retail and leasing strategy decisions.
- Built an Agent-Based Model in NetLogo to simulate workplace social distancing and return-to-office strategies during COVID-19.
- Built UK-wide consumer segmentation model using 10,000+ survey responses and multiple unsupervised learning techniques; widely adopted by commercial teams.
- Designed ETL automation (Alteryx + SQL) reducing monthly reporting effort for 40+ shopping malls from 1 month (3 people) → 1 day (1 person) with drastically improved accuracy.
- Engineered a fully automated ETL pipeline for a national benchmark product, enabling fast ingestion of survey data and powering interactive Qlik dashboards.
- Created real-time sentiment dashboards (Python + Twitter API + R Shiny) to analyse customer emotions across UK shopping centres.
Education
MSc – Smart Cities and Urban Analytics
2017 - 2018
University College London, UK
BEng – Civil Engineering
2013 - 2017
University of Shanghai for Science and Technology, China
Personal Projects
Data Insight Scientist
2021 - 2022
COMMUN Ltd – London, UK
- Designed GCP-based ETL pipelines and analytics architecture to support public-sector grant decision-making.
- Developed dashboards and NLP models to analyse community funding patterns across the UK.
Portfolio
A mix of applied analytics, experimentation, and playful side projects – from mobility and offer prediction models to a personality-driven Streamlit app, “Penguin Love Judger.”
Universal Queue AI – Theme Park Wait-Time Simulator
A machine learning powered queue simulation tool built for Universal Studios Florida. It predicts ride wait times using historical queue data, weather patterns, and crowd dynamics — helping visitors plan smarter, avoid peak congestion, and maximise time in the park. A fun blend of operations research, forecasting, and real-world experimentation.
Penguin Love Judge
A playful Streamlit “relationship moderator” that turns couples’ answers into cheeky verdicts and tailored suggestions on how to fight better, communicate clearer, and strengthen their relationship.
Define Footfall Using GPS Data
Large-scale mobility analytics using GPS and POI data to derive footfall, catchment areas, trip purposes, and support leasing and retail network decisions.
Starbucks Offer Prediction
Offer response and uplift modeling using customer transaction history to predict which users are most likely to engage with specific promotions and journeys.