EXPERIENCE
TARGET, Seattle, US
| Lead AI Scientist | Jan 2022 - Present |
AMAZON, Seattle, US
| Data Scientist II | July 2017 - Jan 2021 |
[1] Built and deployed an ensemble regression model in AWS to forecast all US IMDb TV-series titles’ performance (e.g. – streams, hours, customers) on Amazon AVOD video streaming service to support content acquisition. Improved WMAPE from 55.7% to 39.6%. Enabled stakeholders to target key demographics by further breaking down title projections by age, gender etc
[2] Developed forecasting model for Retail’s bottom-line cash flow and its drivers at product line level using Prophet in Python. Delivered insights into supply-chain levers driving cash flow trend breaks (e.g. – shipping speed, selection changes, subscription fee) while improving MAPE by 724 bps over naïve projections previously used by Finance partners
[3] Collaborated cross-functionally to parameter-tune forecasting model for fringe items’ demand (84% of Amazon’s selection) using a two stage model with MLP Class Probability Estimation & Quantile Regression to predict demand distribution using SK-Learn in Python. Improved P90 Quantile Loss by avg. 1737 bps for CA/BR-Retail and IN-FBA
[4] Developed anomaly detection model using distance to holiday and catalogue features to monitor impact on forecast error from model launches, seasonality, deals, item mix or YOY growth; delivered leaders deep insights on business health
[5] Proposed and implemented a white paper for an SVP level metric of retail out of stock (ROOS) caused by forecast under-bias and built an attribution model of root causes – unhealthy inventory markdowns, late/unsupported promotions, incorrect seasonality/lifecycle stage & other drivers to triage 3.6M weekly ROOS impressions
[6] Identified existing forecast gaps through deep-diving top penalty anecdotes and performed opportunity analyses to drive new feature development for – hardlines’ preorders, moving holidays, halo effect from movies, fraudulent demand etc
OPERA SOLUTIONS, Noida, India
| Research & Data Specialist II | December 2015 – May 2016 |
[1] Client: 4 US based Hospital Networks : Applied machine learning using Python on patients’ invoice data to predict missing charges; reduced time and inaccuracies in reviewing patients’ bills by assisting client’s auditors with an avg. monthly net impact of $200k per client
GITHUB </>
[1] Anti Money Laundering (AML) / Fraud Detection [Link]
[2] Built an NLP model of vocab/semantics to detect sarcasm in news headlines with Python, NLTK & Word2Vec [Link]
[3] Time Series Forecasting with Exogenous Features [Link]
[4] Tennis player recognizer with resnet34 CNN architecture in Fast.Ai and transfer learning [Link]
[5] Performed real time analysis of players’ popularity using Twitter streaming API, PySpark, Python and Plotly [Link]
[6] Built a football news aggregator by web-scraping articles’ titles, dates and links from multiple sports websites [Link]
[7] RShiny App for Video Games / Movies / TV Shows [Link]
[8] Deep Learning [Link]
[9] Recommender System for Movies with Collaborative Filtering, Matrix Factorization & Deep Learning [Link]
EDUCATION
UNIVERSITY OF MINNESOTA, Minneapolis, MN
| Master of Science, Business Analytics | May 2017 |
BITS PILANI, Goa, India
| Bachelor of Engineering, Mechanical | July 2013 |
GRADUATE PROJECTS
CARLSON ANALYTICS LAB, University of Minnesota, US
[1] Travel Recommender System – Experiential Learning Live Case
Collaborated with a cross-functional team of 5; personalized the search results of a hotel booking website
Built collaborative filtering and content based recommender systems in GraphLab; presented client an RShiny UI
[2] Forecasting for B2B Travel Firm – Experiential Learning Live Case
Performed time series modeling in R to forecast bookings volume; reduced mean absolute percentage error by 6.3%
[3] News Articles Classification & Sentiment Analysis
Built NLP model to classify news articles through ML-PySpark; extracted sentiment score and n-grams to detect events
CERTIFICATIONS
[1] Deep Learning Specialization
- Neural Networks and Deep Learning
- Improving Deep Neural Networks: Hyperparameter Tuning, Regularization and Optimization
- Structuring Machine Learning Projects
- Convolutional Neural Networks
- Sequence Models
[2] DevOps on AWS: Code, Build, and Test
[3] DevOps on AWS: Release and Deploy
[4] DevOps on AWS: Operate and Monitor
[5] SAS Certified Base Programmer for SAS 9
[6] Python OOP - Object Oriented Programming for Beginners
[7] Apache Spark with Scala - Hands On with Big Data!
[8] PyTorch: Deep Learning and Artificial Intelligence
[9] Building Recommender Systems with Machine Learning and AI
SKILLS
- Supervised ML - Regression, Classification, Forecasting, Recommender Systems
- Unsupervised ML - Clustering, Anomaly Detection, Similarity Matching
- Statistical Inference - Estimation, Hypothesis Testing, Explanatory Modeling
- Metrics & Controlled Experiments
- Deep Learning
TOOLS
- Data Science - Python, R, Scala, Hadoop, Spark, SQL, Hive, RShiny, PyTorch, Scikit-Learn
- DevOps - Unix, AWS (Redshift, Lambda, EC2, Sagemaker, Step Function, Cloud Formation), Oozie