TechAscent Blog

Benchmarking some solutions to a common problem.

A well-written, *FAST* data science database for your laptop.

Regarding a simple way to produce parquet files in Clojure.

Getting started doing data science in Java with tech.ml.dataset.

We bring a simple and general foreign function architecture to life using Clojure.

We learn all about memory mapping, its history, and the Apache Arrow 1.0 binary format.

Our business-focused strategy for realizing value from early stage prototypes.

We follow a function object from its Clojure definition to its call from libpython.

Building responsive applications with great SEO and mobile performance often means using a combination of both server and client side rendering.

We talk about JNA, a (somewhat) hidden gem in the JVM ecosystem.

Taking advantage of modern software means using all the tools available on your system. We present a simple, helpful tool for working with non-JVM entities.

A simple scheme for managing features and data with clients.

We love empowering experts; doing so is the most rewarding part of our job. It is fundamental to the TechAscent approach and business.

TVM is a compiler stack for deep learning systems. Here we demonstrate it with a bespoke data intensive algorithm finding both correctness and performance far simpler to achieve.

Writing high performance code is difficult. Some advances in the field of numeric computing can help significantly.

An example of leverage. We love OpenCV and want it to play a bigger part Clojure ecosystem. We use concepts from earlier posts to make interacting with OpenCV quick and painless.

Building bindings to libraries outside the Java ecosystem allows us to adapt to changing technical demands. An intelligent approach here reduces the cost of developing bindings and multiplies the resulting leverage.

Performant code often depends on efficient manipulation of contiguous arrays of numbers of different types. Our datatype library makes this fast for computers and programmers.