Data Scientist
Resume:
Hamza Humayun
********@*****.***
https://github.com/hhumayun
SUMMARY
Full-stack Data Scientist with over 5 years of experience in deploying innovative data solutions to tackle business problems. Comfortable structuring solutions to ambiguous problems. Research-oriented with deep knowledge of statistics and machine learning algorithms.
Deployed predictive model that doubled the ROI on telemarketing campaigns for a major Canadian charity
Conducted the first ever Digital Attribution study in the automotive industry that included data from both online digital analytics and offline purchases. Developed website users’ segmentations and personas were as part of project
Developed sales forecasting solution for large automotive client, outperforming client’s internal model by 50%
Increased click-rate by 40% for major grocery manufacturer by developing a recommendation engine
Led team that deployed an end-to-end crime mapping solution for the police in one of Pakistan’s largest cities. Initial pilots indicate a decrease in overall crime of 30%
EXPERIENCE
Co-founder Aug 2018 – Mar 2019
Insurance Technology Startup
Co-founded startup providing a platform for insurance companies to implement activity-based rewards programs to improve population health
Contributed to product design features. These include news feed, motivational messaging, setting up of fitness challenges and analytics to be provided to user and insurance companies.
Designed and implemented the product using a reactive backend microservices infrastructure in Scala based upon the Lagom framework and Cassandra. As part of backend also implemented a streaming service to pull data from FitBit’s Web API.
Developed a simulation framework to help set points per activity and the value per point
Data Science Consultant Aug 2018 – Nov 2018
Engage Consulting, Working Remotely (Part time)
Advising on overall statistical, data and technological aspects of the business including data collection methods, survey design, fraud detection and statistical evaluation of product effectiveness
Performed candidate segmentation to identify different types of candidates across different variables
Developed Bayesian hierarchical model to estimate effects of university, major and sex
Data Scientist Feb 2017 – Aug 2018
DDB Canada, Toronto, Canada
Data science subject matter expert consulting for major internal and external enterprise clients. Provided advice across all internal initiatives related to data science/data engineering. Interfaced with clients to understand problems and design and implement solutions. Mentored members of analytics team on predictive analytics and how it could be used to benefit their projects
Digital Attribution Analysis
Conducted digital attribution study for Volkswagen, which was the first digital attribution analysis for a Canadian automotive company that included both online and offline purchases
Developed segmentation / personas of web behavior using machine learning algorithms. This segmentation was further enhanced using data from Environics Census, Opticks and PRIZM5 datasets
Linking offline sales to online behavior was accomplished using innovative heuristics leveraging off of existing data processes, requiring almost minimal resources to set up the data process.
Analysis was conducted using Spark (Scala) and Hadoop and H20 by analyzing hundreds of millions of website logs (over 300 GB in size)
Donation Increase Prediction
Modeled donor behavior to identify those likely to increase monthly contributions if targeted. Telemarketing campaign based on output of model led to a doubling of success rate per call
In addition to client data, increased prediction accuracy by integrating exogenous data sources as Environics PRIZM5 segmentation and Environics enhanced census
Zero-hassle deployment at client site. Solution seamlessly integrated with client’s existing data structure and systems, allowing them to plug data directly from their existing data sources
Sales Forecasting
Deployed sales forecasting solution for large automotive outperforming client’s internal model by over 50%. Model is in production and used to decide if sales campaign should be launched to meet sales targets
Time to next purchase modelling
Developed model for Volkswagen to predict an individual’s time to next purchase. Targeting based on model led to increased marketing effectiveness across all channels
Project involved sophisticated data modelling and engineering challenges requiring aggregation and standardization of data across heterogenous datasets
Integrated exogenous data sources such as economic data, Environics PRIZM5, Opticks and Enhanced Census datasets to increase prediction accuracy
Modelling involved writing custom Hidden Markov Model that would scale to size of dataset
Recipe Recommendation Engine
Developed recipe recommendation engine for Kraft using Spark MLIB that increased click through rate by 40%. Service implemented using Scala, Kafka and Spark Streaming to provide real time recommendations for 1M daily website visits
Build and Price Completions vs Sales
Developed Hierarchical Bayesian regression model in PyMC3 and Stan to uncover latent relationships between online build & price activity and sales by model and month.
Used by management to calibrate spend on Build and Price ads
Hierarchical model helped make robust inferences from limited data available by sharing statistical strength across car models and months
Excel Solutions Consultant / Data Science Prototype Developer Sep 2016 – Feb 2017
Vena Solutions, Toronto, Canada
Interfaced with enterprise clients to understand business problems and develop excel-based budgeting/forecasting solutions in Excel by combining VBA macros and Vena’s core cloud CPM product
Strategizing and implementing potential cloud data science services that Vena could provide based on customer data stored in Vena’s cloud service. Prototypes included include anomaly detection in accounting data, segmentation and demand forecasting
Data Science Manager May 2013 –Mar 2016
Technology for the People Initiative, Lahore, Pakistan
Managed a team of 5 data analysts and web developers to solve data problems for the government. Liaised with officials in different departments to understand problems, provide advice and design and deploy solutions
Spatial Analysis of Crime Data
Led team that collaborated with the Police in a city of over 10M to identify spatial and temporal crime hotspots. Deployments based on model led to a significant decrease in crime in areas where was tested on the ground
Project included designing complete solution from data collection, analysis/modelling, application development for visualization and deployment
Crime hotspot patterns were identified using custom Kernel Density Estimation clustering algorithm written in Matlab (see GitHub for code)
Final data product delivered as a web application, visualizing crime points and hot spots based on clustering algorithm
Other Projects
Analysis of over 80GB of cellular usage data using Apache Spark resulting in a paper in ICT4D that established value of cellular data as a proxy for socio-economic indicators
Developed predictive model to classify citizen complaints sent via SMS as ‘bribe demanded’ and ‘no-bribe demanded’.
Actuarial Associate Pricing Jan 2012 –Mar 2013
Reinsurance Group of America, Toronto, Canada
Managed group LTD, life and out of country medical re-insurance pricing
Took initiative to convert LTD pricing models from Excel to AXIS for consistent and faster pricing
Using Excel and VBA set up template to perform termination experience study on a portion of group LTD business to identify factors affecting profitability of inforce business
Actuarial Associate Pricing Oct 2008 – July 2010
Manulife Financial, Affinity Markets, Toronto, Canada
Managed overall pricing process of Extended Health & Dental plans. This included analyzing data for abnormal claim activities, checking assumptions, developing claims costs, liaising with marketing to price benefit enhancements and developing and publishing rates.
Conducted Lapse experience study for the entire business unit
Conducted research on how to price new product features
Actuarial Analyst Nov 2006 – Oct 2008
Towers Watson, Detroit, USA
Responsibilities included performing valuations to determine pension plans’ liabilities producing reports, drafting memos and communicating with client teams to ensure smooth workflow. Worked on numerous special projects such as allocation of plan assets after spin-offs, accounting impact of plant shutdowns & modification of benefits, and projections under FAS 87 and IAS 19
SKILLS AND TECHOLOGIES
Technologies
Python
R
Julia
Excel & VBA
Scala
Tensorflow
Stan/PyMC3
Spark/Hadoop
SQL
Cassandra
H20
Concepts
GLMs
Classification
Deep Learning
Probabilistic Graphic Models
Bayesian Inference
Actuarial Pricing
Probabilistic Programming
Clustering / Segmentation
Time series forecasting
Natural Language Processing
Recommender Systems
Domain Driven Design
Causal Inference
Reactive programming
EDUCATION
University of Waterloo
Bachelor of Mathematics Joint (Honours) Computer Science and Actuarial Science
Society of Society of Actuaries
All exams written for Associate of Society of Actuaries (P, FM, M, C and FAP Modules)
Contact this candidate