Jaxon — Eliminating The Biggest Bottleneck In AI
Automating the process of labeling data, allowing data science teams to quickly build fully-trained models.
Artificial intelligence is currently powered by armies of humans. Machines cannot ‘read’ natural language like humans do and require labeled data to ‘learn’ (adjust neural network weights). Massive efforts are underway to manually label data; typically hundreds of thousands of examples are needed to train machine learning models, and even more for deep learning. Human-powered approaches are extremely slow and can take months, be expensive, wildly inconsistent, biased, and error-prone (often three+ humans are put on the same labeling task to increase accuracy).
What The Company Does
Data is fed into Jaxon through an elaborate assembly line that dissects, analyzes, and then refines itself on the fly. Jaxon dives into the data, discovers correlations, and looks for patterns and connections. Dozens of algorithms work together to assign the best possible labels to the original data. With Jaxon, data labeling is done in hours versus months. Jaxon is a powerful technology enabler for any organization that is applying artificial intelligence to text. The company is very attractive in situations where human labelers struggle with or simply can’t provide adequate support due to the sensitivity of the data. With push-button functionality, non-data scientists can use Jaxon to create custom classifiers on demand, opening up a wide range of opportunities.
The AI market is expected to exceed more than $191B by 2024 at a CAGR of 37% in the given forecast period, according to Market Research Engine. An estimated 10% to 15% of the cost of building AI is data labeling and creating training pipelines.
Data analysts, engineers, and scientists will use Jaxon. They will pay for access to the studio, the ability to create custom labelers, and gain access to The Collection, a walled garden of metadata. Enterprise customers pay for annual subscriptions. The company has a stratified pricing model tied to the number of concurrent “projects” they want to be able to access. It starts at $25K per year for one concurrent project. It costs $75K/year for five concurrent projects, $200K for 20, and then is priced custom from there, depending on how many project slots are needed. Jaxon is also now live on AWS Marketplace.
Jaxon is in talks with more than 200 prospects by leveraging its networks and going direct via email and InMail. The company is cultivating relationships with large system integrators, resellers, and referral partners for channel sales and teaming on delivery, but will remain focused on direct sales until sufficient traction is reached. Jaxon also secured a contract with the U.S. Air Force in which they will partner with MIT, who will leverage Jaxon’s platform for object detection in images. The company has launched a bifurcated inbound and outbound strategy, is presenting at trade shows virtually, and starting to push on social and traditional media outlets. The technology is domain agnostic, but is focused on certain types of data and focused on use cases that are applicable horizontally. Jaxon has been targeting retail, insurance, financial services, and national security.
Founding Team Background
Jaxon has a proven and experienced team behind it. This is the third company that Jaxon’s co-founders have founded together — the previous two companies were grown and sold. Jaxon’s EVP of Sales, Paul Reston, has taken a company from scratch to a $1.3B division of IBM. Jaxon’s core team has been together for several years now, successfully building the previous company BigR.io into a big data and ML consulting firm for the Fortune 500 before focusing on Jaxon full-time.
The company is currently raising its second seed round and looking for angel investors. Connect With The Jaxon Team.
Subscribe To Storied To Get More Startups In Your Inbox