Profile Summary
More than 6 years of data science experience include 2 years of leading high performing teams. Experienced insolving complex ML + data problems and clear communications with stakeholders and leadership

Role 1 : Lead Data Scientist
- Deployed ARIMAX based time series model for predicting google ads campaigns ROI parameters at daily level as part of a product offering, with 20% improvement in ROI over 2-month test period.
● Deployed keyword clustering model (topic modelling) to generate brand specific keywords for improved performance, with amazon ads
campaign performance uplift (Cost per acquisition) of over 35% w.r.t manual campaigns
● Reduced Cost per acquisition by 50% using Mutli-Touch Attribution based cross-platform budget optimizer tool
● Developed a bid optimizer using XGBoost based multi-class model to drive higher margins for an ad serving platform
Role 2 : Data Scientist Products/Data Science Solutions Developed
- Assortment Optimization – Recommends optimal store assortment for general merchandise (low purchase frequency, high churn) item groups. Attribute based solution leveraging genetic algorithms to effectively learn from sparse data. Potential savings of > 5 million USD per category through improved decisions. Scaling using PySpark
- Auto linkage – Automated solution for classifying items into groups based on item attributes. Supervised learning using one class SVM used along with K-nearest neighbor algorithm for classification
- Digital merchant – Designed automated system for anomaly detection and root cause identification for merchandising KPIs(inventory level, sales, logistics cost etc.). System capable of identifying short and long term trends. Ensemble of algorithms (~Bayesian change point analysis, decision trees, cumulative gradients) used.
- Developed product to calculate price elasticity for more than 5,000 items belonging to 37 product categories for wholesale purchasing environment Big data scaling implemented through Hadoop streaming
- Developed dynamic programming based solution for pricing managers to recommend phase wise pricing strategy for clearance markdowns.
Role 3: Data Scientist, Marketplace Team
- Deployed machine learning models in production including data pipelines, feature creation, automated model training, model serving and monitoring
- Designed and implemented monitoring + alerting systems to keep track of health of production models using tools like BigQuery, Data Studios, Airflow
- Evaluated impact of new product rollout by designing randomized control experiments, hypothesis testing and presenting results to stakeholders
- Reduced the bandwidth used for ad-hoc tasks by 66% through activities like code optimizations, dashboard revamp and standardising analysis process
- Served as the Data Science representative in marketplace architecture group where I was responsible for good design & architecture of tech products
Skills
Artificial Intelligence & Machine Learning, XGBoost, Random Forest, Neural Network, R, Python, C++, Spark, Hadoop, Hive, SQL, Pandas, NumPy, Scikit-learn, Keras, TensorFlow, CNN, RNN, Natural Language Processing (NLP), Big Data, Agile, JIRA, Robotics process automation