Skip to content


Python and distributed machine learning

In today's computing world, machine learning is hitting a performance block. More and more companies want to run them on-demand, instead of as batch processes and want their ML models to deliver results in real-time. Often, the datasets are big-data. Thus, the ML frameworks that data scientsits learnt (pandas, scikit-learn, pyTorch, keras) and know to use don't scale well in this fast production environment or is too cumbersome to implement. This blog explores the approaches the ML, DevOps, HPC industry has arrived at in 2019.

Design principles behind RESTful APIs

RESTful stands for "Representational State Transfer". Consider this as a concept and a pattern of building client-server APIs. I have been building Python APIs that consume some popular RESTful APIs for the past 5+ years. This article outlines the aspects of a thoughtful and well-designed REST API. Some of the aspects here are from the perspective of a consumer, not the maker of RESTful APIs.