Home / Blog

Jaxon blog

Blog

Product posts, company news and explainers on how AI works.

Jaxon AI prediction of seafloor habitat segmentation — colored polygon labels over coral reef imagery

March 24, 2023

Mapping Seafloor Habitats with AI

Jaxon created a custom AI model to characterize the seafloor habitats of the U.S. Caribbean and accelerated the annotation process from minutes to milliseconds per image. Jaxon incorporated

Read the post →

October 3, 2022

Synthetic Data: Noise vs. Quantity

Greg Harman, Jaxon's CTO, shares his thoughts on improving models with synthetic data, and striking a balance between noise and quantity.

Read the post →

July 28, 2022

Rapid Prototyping Webinar

Does there have to be a tradeoff between speed and accuracy? With rapid prototyping, you can have both. The Jaxon team discusses how.

Read the post →

Conjoined nested loops diagram — task, data, and model loops in iterative AI development

June 14, 2022

Iterative AI Development: Levels of the Loop

Jaxon's CTO, Greg Harman, takes you through the loops of iterative AI development and how to take AI from toy problems to practical applications.

Read the post →

Data science illustration — lightbulb with chart and flask representing experimentation and insight

April 23, 2022

Data Science actually IS ‘Science’

Do you feel like people are treating data science more like 'astrology for data'? Jaxon wants to trade the mysticism for concrete facts.

Read the post →

February 17, 2022

Voice of the Customer Webinar

We discuss using custom ML models to analyze voice of the customer on social media, call/chat logs, reviews, surveys, and more.

Read the post →

May 6, 2021

Ground Truth Multilabel Decision Thresholds

Jaxon's CTO, Greg Harman, explores the meaning of truth—specifically, ground truth and its intersection with real-world data.

Read the post →

Mute swan animation — visual reference for rare-class examples in active learning

March 30, 2021

The Trouble with Active Learning

Active learning was an important step towards creating effective TDP and cost-sensitive ML, but is it enough by itself?

Read the post →

March 12, 2021

Call Transcript Classifiers Webinar

Jaxon's approach to call transcript classifiers is a 3-component neural model: text representation, attention layer, and classification.

Read the post →

Camel and longhorn cow in a dry field — visual metaphor for covariate drift between training and production data

September 2, 2020

Covariate Drift & SmartSplit Technology

How can Jaxon's SmartSplit function help you avoid covariate drift and split off test sets smarter? Jaxon's CTO, Greg Harman, explains.

Read the post →

August 23, 2020

SmartSplit Technology Webinar

Jaxon’s patent-pending SmartSplit Technology is a proprietary means of splitting a dataset (e.g. into training and holdout datasets).

Read the post →

July 21, 2020

Data Augmentation Webinar

In this webinar, we explore questions like, What is data augmentation, how does it work, and why is it so much harder to augment text than images?

Read the post →