Technical Products

Privacy and ownership of your own data is increasingly being peddled both as a regulatory issue as well as a product issue. Just don't be fooled: it's plenty possible to uniquely identify your device without asking for consent. Here's how.

Have you always wanted a blog just like The Damn Blog? Fear no more: this easy tutorial will walk you through the entire process, from start to finish, in 7 very simple steps!

Most people can 'easily' understand and visualize up to 2 dimensions, so very often - when visualizing networks of similarities - you will consciously or unconsciously try to reduce down to two dimensions. Here's a few ways to go about that.

Facebook has announced its own Cryptocurrency: Libra. I want to touch on a couple points which I believe the current conversation lacks: its monetary and economics implication, its role in developing economies, and its use as a payment method.

Smartphone photography is bridging the gap between small-sensor cameras and full frame sensors thanks to much smarter post-processing. Such a powerful technology in our pocket all the time - but can we replicate this effect at home?

As more and more people get interested in data science, the risk of getting biased insights by applying cookie-cutter models is definitely on the rise. But what we're interested in is cause, not spurious correlation. How can we find causal relationships in the real world?

My blog posts are plagued with typos - an unfortunate side effect of writing at night. So I can either use a spelling checker, or write one - but how does one write a spelling checker actually?

Does regulation do anything good for airplane accidents? We can find an answer to this question - and analyze any moving process to see if it has changed over time - using Bayesian statistics in a fully probabilistical framework.

What makes a cat, a cat? Using unsupervised methods, we can create a snappy visual search engine by implicitly teaching our model how a kitten looks like.

Given a number of potential returns associated with certain tasks, and given limited allocable time for multiple people, how can we make sure everyone's time is used to the max for the best global welfare?

Complex models can be hard to distribute, host and serve. Luckily we can use Docker to deploy a state-of-the-art text classifier from start to finish.

A SAT solver can be a brilliant way to satisfy multi-constraint optimization problems. But converting logical constraints to the right format requires some brain muscles!

Building a generative Variational Autoencoder from scratch to imagine Fashion MNIST clothes from a sampled Gaussian latent space: who said AIs can't be creative?