Our Latest Blog Posts

Using DC.jS and crossfilter with React

A technique to get interactive high dimensional charts in React

When Should You Use Machine In The Loop

Machine in the loop techniques reduce the cost of annotation but introduce bias. We discuss how to manage that tradeoff.

Character Level NLP

Character level models for NLP facilitate large vocabularies and remove computational bottlenecks during model training. This post reviews those traits and discusses the drawbacks of character level models for NLP.

Understanding the Tensorflow Estimator API

We explain the motivation behind Tensorflow's Exstimator API, the problems it solves and how to use it

Efficiently Labeling Data for NLP

Three tips and tricks to get more labeled data with less work

Embrace the noise: A case study of text annotation for medical imaging

How a company used NLP to annotate chest X-rays and what we can learn

Active Learning: Optimization != Improvement

Active Learning promises to reduce the labeling cost for building a model. We look into how that works out and ask if active learning delivers on what we really need

Bug Postmortem: Wrong image deployed on Docker Swarm

Postmortem analysis of a production bug where a customer as shown the wrong version of our frontend