Tech Lead, Ads Economics Platform

Supporting Ads ML Infrastructure use-cases

Technologies used: Kotlin/JVM based Microservices, Python, Flink, Spark, Kubernetes

Team: Alexa AI Machine Learning Platforms

► Architecting large scale distributed systems for Alexa machine learning use-cases.

► Designing microservices for managing 2900+ AWS accounts and maintaining infrastructure through code.

► Collaborating with scientists and engineers to evaluate design approaches, prototype new technology, and determine technical feasibility.

Relevant Projects

► Built platform to collect various infrastructure usage metrics to produce dashboards for both Operational Platform Visibility and Business Analytics.

Technologies Used: Lambda (Serverless), Kinesis, CloudFormation, Redshift, Apache Superset

► Revamped Production NLP Alexa Model in TensorFlow to achieve faster training using Multi-GPU Distributed Training (Horovod) and enhancing data pipelines.

► Contributed to project creating a Python API for internal C++ based Machine Learning framework

► Engineered a Python data pipeline package that handles variable transformations,

aggregations, and cleaning for dataset creation required in predictor variable

identification for credit decisioning model

► Responsible for creating materialized views for delving into metrics obtained from

digital channels to drive higher customer engagement, NPS, and other KPIs

► Launched a data visualization dashboard using D3,js, Angular.js, Flask, and

Teradata SQL; revitalizing the existing Tableau version

► Overhauled existing UK credit bureau data monitoring scripts from SAS to R,

involving comparative analysis of historic data and ensuring that recent data

follows historic trends. Output reports were generated using Knitr and ggplot2