Iliyan Zarov

myLevels uses data from continuous glucose monitors and Bayesian machine learning to model the very personalized impact that food has on people’s bodies. We help you see and understand what foods are best for your unique metabolism. People using myLevels for weight loss have dropped more than 10 kilograms.

We consider the problem of inference in a causal generative model where the set of available observations differs between data instances. We show how combining samples drawn from the graphical model with an appropriate masking function makes it possible to train a single neural network to approximate all the corresponding conditional marginal distributions and thus amortize the cost of inference. We further demonstrate that the efficiency of importance sampling may be improved by basing proposals on the output of the neural network. We also outline how the same network can be used to generate samples from an approximate joint posterior via a chain decomposition of the graph.

Evoqus is a cloud platform delivering effortless scalability for technical computing in Julia. Code can be executed with a single command across hundreds of CPUs and GPUs in the cloud.

The stateless back end services are written in Clojure. The platform can be accessed using the CLI, written in Go, or the web app built on React/Om and ClojureScript.

Evoqus grew out of a side project aiming to make it extremely easy to move the computing from my laptop to the cloud on demand. The details of the underlying cloud instances, storage, and network are transparent to the end user while retaining full control over the workers and the environment.