Turning data into decisions

Behind almost every important decision is a mountain of data. Flood management, infrastructure planning, mineral exploration and environmental monitoring all rely on information collected over years, sometimes decades. Yet much of that data remains underused because analysing it requires significant technical expertise.

University of Auckland alumni Ashin Alex, Sam Kurian and Jimin Seo believe the challenge is not just in collecting information, but in making sense of it.

Through their startup Hyades, the trio are developing technology designed to help organisations turn complex datasets into predictions that support real world decisions. The technology is particularly focused on spatial data – information tied to a physical location.

“Companies come with the data, and they come with a question they want answered from that data,” says Jimin.

That question might be whether a river is likely to overflow following heavy rainfall, where a mining company should focus future exploration efforts, or which areas are most vulnerable to flooding.

The information needed to answer those questions often already exists. Councils, researchers, businesses and government agencies collect extensive environmental and geographic data: everything from rainfall records and river measurements to satellite imagery and geological surveys. The difficulty lies in bringing that information together.

Datasets are often stored in different formats, collected using different methods and designed for different purposes. Before any prediction can be made, the data must be cleaned, organised and transformed into a format suitable for machine-learning models.

Traditionally, that process has required specialist expertise to synthesise these datasets into a unified format that is ready for machine-learning training. Hyades aims to automate much of that work.

“We’re essentially providing the capability of an added data engineer, machine learning engineer and an analyst in one,” says Ashin.

Using the company’s platform, organisations upload their data and describe, in plain language, what they want to learn from it. The platform then analyses the information, prepares it for modelling and tests different approaches to determine which model is best suited to the task.

In simple terms, it acts as a bridge between raw information and useful predictions. 

Unlike generative AI tools that can produce convincing but inaccurate responses, Hyades is designed around transparency and verification. Rather than simply generating an answer, the platform shows users how a model has been created, allowing them to review and evaluate each step. The goal, says Jimin, is to keep humans in the loop and give organisations confidence.

For co-founder and software lead Sam Kurian, the approach grew out of a frustration familiar to many researchers and data scientists. Building machine learning models is often a repetitive and time consuming process requiring many technical decisions along the way.

Sam began exploring ways to automate those steps, creating tools that could test and optimise models automatically rather than relying on engineers to perform every task manually. That early work eventually evolved into a platform capable of handling much larger and more complex datasets.

Ashin, whose background is in physics and space systems, describes the company’s goal as making spatial intelligence easier for non-specialists to us independently.

“Our vision is to embed spatial intelligence everywhere,” he says. “We’re trying to enable extra capability for organisations to make much better decisions.”

Importantly, the work does not stop once a model is built. For example, a flood-risk model created several years ago may not account for changing weather patterns. Environmental conditions shift, new data becomes available and assumptions made during the original modelling process can gradually become outdated.

Hyades’ platform is designed to continually monitor and improve models as new information becomes available. This gives customers greater ownership of th process, and the opportunity to benefit from it long term. Ultimately, access is expected to operate as a subscription model. For the founders, this ongoing process is as important as creating the initial model itself.

The company has already worked across a variety of use cases, including environmental monitoring, natural hazard forecasting and research projects requiring complex predictive analysis.

While the technology itself is sophisticated, the founders see their mission in relatively simple terms: helping organisations make better use of information they already have.

Hyades’ long-term vision is to help unlock the productivity and capability of people who are not data scientists or machine-learning engineers, allowing more organisations to benefit from data-driven decision making.

That vision has been shaped by the different backgrounds each founder brings to the company.

Ashin contributes expertise in science and machine learning. Sam leads the software engineering behind the platform. Jimin, who studied law and global studies, focuses on partnerships, operations and understanding customer needs.

Together, they represent an unusual, but complementary, mix of disciplines.

“I can bring clients on board and handle a lot of the operational work, knowing that Ash and Sam are doing fantastic R&D and tech development work and vice versa,” says Jimin.

While their academic backgrounds differ, all three founders credit the University environment with helping them explore ideas beyond their own disciplines. For Ashin and Sam in particular, the Faculty of Science provided opportunities to work across research, software development and emerging technologies, while also connecting with other students from very different fields. Those intersections ultimately helped shape the company they would go on to build together.

The combination reflects a broader truth about emerging technologies. Building successful AI systems is not only a technical challenge. It also requires an understanding of the problems people are trying to solve.

For Hyades, the goal is not to create more data. It is to help organisations turn existing data into decisions. And in a world already overflowing with information, that may be the more significant challenge.