An AI company builds and maintains a sophisticated text-to-speech system. The archetypal problem is deciding how to pronounce a word when it has multiple options — should “read” be pronounced “red” or “reed’.
They have a team that check the results and provide feedback to teach the AI system. However, translating this feedback into something the AI can learn from is difficult.
The company aims to solve this by enabling the team to view network graphs that represent the decisions the AI made and then correct the decisions by modifying the graphs. This is an intuitive and powerful way to enable non-technical team members to operate in the language the AI understands — structured graph data.
Instead of spending 6 months building a custom tool for this, the company’s engineers can build the tool in an afternoon and immediately deploy it to their team. The engineers focus purely on the logic of their graphs, not on the nitty-gritty of visualising and modifying graphs. As the team uses it, the engineers can improve it quickly while spending most of their time on their core focus, improving their AI.
A data-science consultancy creates network graphs mapping interesting data such as media trends, topic-based influence networks, and ecosystem stakeholder networks, for their clients. They provide an end-to-end service, but getting the resulting graph just right involves a lot of ad-hoc back and forth. By using Neuclid they are able to develop a more powerful interactive tool that enables their clients to immediately begin investigating the graph data as the consultancy compiles it.
Their interactive tool allows clients to select which aspects of the graph to focus on and what properties to highlight. They can perform custom calculations and computations on the graph, such as influence metrics on specific topics for specific stakeholders.