From data to decisions: building AI that works in the real world
Professor Daswin De Silva is leading applied AI research at La Trobe University that turns complex data into practical systems – from energy optimisation to industry training – with a focus on efficiency, responsibility, and real-world impact.
Artificial intelligence (AI) has moved quickly from novelty to infrastructure. It now underpins how organisations manage energy, plan operations, and make decisions. That shift has exposed a new challenge: ensuring AI systems are reliable, efficient, and genuinely fit for real-world use. At La Trobe University’s Centre for Data Analytics and Cognition (CDAC), Professor Daswin De Silva is focused on meeting that challenge head-on.
Prof De Silva is Discipline Lead of AI and Analytics, Director of AI Strategy, and Deputy Director of CDAC. Across these roles, he leads research spanning machine learning, cognitive AI, energy systems, and workforce training, with a consistent emphasis on application rather than theory.
“We build and apply,” he says. “We work on building new AI algorithms with novel capabilities, and applying AI algorithms to real-world problems that deliver meaningful impact to industry, society, and the economy.”
That philosophy shapes CDAC’s work from research design through to industry collaboration.
One of the Centre’s flagship projects is the La Trobe Energy Analytics Platform (LEAP), a campus-scale AI system designed to turn complex energy data into practical decision-making tools. Modern energy systems generate enormous volumes of data, from smart meters and solar generation to battery storage and building-level energy use. Across a multi-campus university, that quickly adds up to billions of data points.
“In LEAP, we’ve developed a robust data lake to collect this information reliably. That data becomes a large, high-quality training set for new AI models,” Prof De Silva explains.
Those models are used to predict energy demand and generation, detect faults and outages, identify underperforming buildings, and optimise battery use to reduce reliance on the grid and respond to peak pricing.
“It matters because we’ve transformed university energy systems into an AI engine that drives informed decisions and actions,” he says. “It’s used daily by the engineering and facilities management teams at La Trobe.”
Andrew Jennings, Executive Director of Net Zero, Facilities, Assets and Services at La Trobe University, began working with Prof De Silva and CDAC in 2020 as part of the University’s Net Zero program.
“A key part of our Net Zero strategy was taking control of La Trobe’s energy data, processing it so it was genuinely useful, and then using AI to generate actionable insights,” Mr Jennings says. “Working with Daswin and the CDAC team, LEAP has delivered well over $250,000 in actual savings for the University, supported more than 15 published research outputs, and given many students access to a real-world environment where they’re solving problems with live data in a growing sector of the economy.”
LEAP’s impact has also been recognised nationally, receiving a TEFMA award for innovation and impact in 2021. More broadly, it demonstrates how universities can host long-horizon, campus-scale AI systems – providing real-world testbeds where complex technologies can be developed, validated, and de-risked before being transferred to industry.
Smaller, smarter and more responsible AI
Alongside LEAP, Prof De Silva leads research into energy-efficient and cognitively inspired AI systems, including approaches such as vector symbolic AI and neuromorphic computing. This work reflects a broader shift across the AI field: simply making models larger is no longer delivering proportional gains, while energy use and operating costs continue to rise.
Instead, his focus is on building smaller, more efficient AI systems that can reason, plan, and support decisions without the heavy resource demands of large models.
“We’re working on energy-efficient models that can achieve comparable performance to much larger systems, and on responsible, safe, and secure AI applications,” he says.
That emphasis on responsibility and practicality extends to CDAC’s industry partnerships. The Centre has built a strong track record of translating academic research into operational systems that deliver value for industry partners while also supporting publications and new funding.
“Our strength is in transforming academic research into practical applications that deliver impact to diverse stakeholders – industry, society and the economy,” Prof De Silva says.
Mr Jennings sees that capability as a key differentiator.
“Daswin has the rare ability to move seamlessly from an academic lens to a commercial one,” he says. “He understands industry drivers and ensures projects are structured around realistic costs, timelines, and deliverables, while still applying cutting-edge AI. That combination is what makes these collaborations work.”
This approach has led to tangible outcomes, including LodgePro – an AI system for lodging tax returns that won an Australian Good Design Gold Award for responsible AI innovation – and national recognition for CDAC’s leadership in teaching AI across postgraduate, undergraduate, and industry cohorts. Through programs such as the OptusU AI micro-credentials, Prof De Silva’s team has supported close to 500 Optus employees to build practical AI skills and confidence.
Collaboration with Innovation Central Melbourne plays an important role in accelerating this work beyond the university.
“Innovation Central Melbourne provides a platform for deep industry engagement, increased research visibility, and employment pathways for our PhD students,” Prof De Silva says.
At its core, Prof De Silva’s work is about making AI useful, trustworthy, and fit for purpose.
“Fundamentally, it’s about informed decision-making and actionable insights that deliver cost savings, emissions reduction, better use of space, and transparency in how energy-related decisions are made,” he says.
Through Innovation Central Melbourne, La Trobe’s applied AI capability connects with industry partners and technology leaders including Cisco, supporting the translation of research platforms like LEAP into real-world systems.
