Enabling reliable, trustworthy data production

Data Quality, Governance, Anomaly Detection, Metadata, and Data Lineage

We built a platform which ensures data quality at a petabyte scale. We validate semantic correctness of data in Hive, validity of individual records in streaming ingestion, timeliness of partitions landing in Airflow, and more.This has many benefits to the business -- verifying data creation logic in ETL's, ensuring business rules from product teams are held, monitoring ongoing experiments from data scientists, and validating data production from machine learning feature/metric generation services. - used by 80+ teams at Lyft- over 18,000 production incidents prevented

Having an extensible Machine Learning platform is critical to many business functions at Amazon. However, the experience and skill to operate this technologies is equally as important.I taught introductory ML classes at Amazon Machine Learning University to engineers and engineering managers. I also guided a group of graduate students through a Data Science project with Amazon Alexa's Data. Met on-campus at UCI, helped guide their solution, while teaching them about proper engineering/machine learning practices.

Conversational Discovery employs voice-based navigation to help customers discover their Alexa feature -- either first-party or third-party. It starts when the customer speaks out an open-ended utterance like "I'm bored" or "Let's Play a game", which kicks off our multi-turn dialog experience (more particularly, a knowledge-based content recommender system).Our team owns a platform which takes traffic from different disambiguation use-cases across Alexa, experiments on it, then hosts the 'current best' experiences in production. We use Reinforcement Learning, Machine Learning, and Search to achieve this.

· Teaching Assistant for Stat89a: Matrices and Graphs in Data Science Developed

Wrote a Data Mining package in Python for consumer banking data.