DiUS

There’s a new DiUS website. On the surface, it’s a much better, clearer way for us to show up on the internet. But getting there was harder than we expected. The breadth of what we do is… a lot. It has to be. The problems our clients bring us don’t come in neat shapes. They cut across design, product, data and engineering, with AI now more visible across all of it. Some are modernisation. Some are greenfields. Sometimes we’re building end to end, sometimes we’re brought in for something very specific. No two are the same, but the way we approach them is. Thoughtful problem solving, and bringing the right combination of skills and expertise to each one. Trying to represent that clearly, without oversimplifying it, was genuinely difficult. At the same time, we had to make sure we reflected the impact AI is having on how we design, build and deliver software. Not as something new for us, but as something that’s now shaping the work more visibly. That shift is real for the teams we work with. They’re being asked to move faster, absorb new ways of working, and still deliver systems that hold up. In that kind of environment, experience matters. Knowing how software actually gets built, where complexity hides, and what stands up in production becomes even more important. Now, the website has caught up to where we already are. If you know us, you’ll recognise it. If you don’t, it’s a good place to start. www.dius.com.au

A little less screen time, a little more soil. The Sydney team got together today for a terrarium workshop building our own tiny ecosystems (and testing who has the steadiest hands). There’s something satisfying about slowing down, getting a bit hands-on, and creating something that doesn’t need a sprint plan to grow. Great way to catch up, meet a few new faces, and leave with something you can keep on your desk that isn’t another coffee cup. Thanks to The Commons for organising 🌿 Let’s see how these go in a few weeks…

We're #hiring a new Senior Software Engineer - Full Stack, with a FE focus in Melbourne, Victoria. Apply today or share this post with your network.

“Add AI” is easy. Making it actually work in production, improve delivery, and hold up over time? That’s the hard part. We’re hiring a Lead Software / AI Engineering Consultant at DiUS (Sydney, hybrid). This is a hands-on role where you’ll design and deliver real AI solutions, and help clients apply AI in ways that improve how they actually ship software. One project might be embedding AI into engineering workflows. The next, building LLM-powered products using RAG or agent frameworks. Same expectation: solve the problem end-to-end, and build something that holds up. What you’ll be doing: 🔧 Design and build AI-enabled systems and experiences 🔧 Lead projects while staying hands-on 🔧 Work directly with clients to shape problems and solutions 🔧 Apply AI across the SDLC to drive better delivery outcomes What works well here: ✅ Deep full-stack experience ✅ Experience building AI solutions (LLMs, RAG, agents) ✅ Comfortable in ambiguity, able to find a way forward ✅ Leads by doing, not delegating If you want to build AI that actually gets used (and lead from the front), you’ll fit right in. Apply now: https://lnkd.in/g7aJqqrs No external recruiters.

Slow product pages don’t just frustrate customers. They cost revenue. Lower search visibility, higher drop-off, fewer conversions. It shows up fast. And at scale, every 100ms matters. In this work with a major retailer, improving product page performance wasn’t a “nice to have.” It was directly tied to commercial outcomes. The challenge wasn’t fixing it once, it was improving performance without putting revenue at risk in the process. No rewrite. No big-bang release. Instead: ✅ introduce a Next.js layer and BFF to decouple the frontend ✅ roll changes out incrementally, not all at once ✅ use feature switches and edge routing to control blast radius ✅ back decisions with performance testing, not gut feel The result: product pages became usable much sooner, performance and SEO improved significantly, and conversions and revenue followed. More importantly, the team now has a way to keep improving safely, while the business keeps trading. If that tension feels familiar, this one’s worth a read. Link in comments.

Most “lead” roles pull you away from the work. You start as an engineer. You end up managing tickets and writing updates. This isn’t that. We’re hiring a Lead Software / Data Engineer at DiUS (Sydney, hybrid). This is a hands-on consulting role where you move between building software and shaping data platforms, depending on the problem. APIs one project, cloud-native pipelines the next. You stay close to the code, work directly with clients, and build systems that actually hold up in production. What you’ll be doing: 🔧 Design and build end-to-end software and data solutions 🔧 Lead projects while staying hands-on 🔧 Work across cloud, modern stacks, and data platforms 🔧 Turn messy problems into scalable, working systems What works well here: ✅ Real experience with data pipelines/platforms ✅ Comfortable in ambiguity, able to find the path forward ✅ Leads by doing, not delegating 👉 Learn more and apply: https://lnkd.in/g4D4kNKa No external recruiters.

When ecommerce performance slips, it shows up fast. Slower product pages mean weaker search visibility, more customer drop-off, and revenue left on the table. In our latest customer success story, DiUS helped a major Australian retailer lift ecommerce performance without a big-bang rewrite. Using Next.js, controlled incremental rollout, edge routing rules, feature switches, a backend-for-frontend layer, caching, and a performance-tested design system, the team improved key product browsing journeys while keeping the business trading. The result: ✅ product pages became usable much sooner ✅ desktop performance improved from 15 to 93 ✅ SEO score reached 100 ✅ increase in conversions and revenue A practical example of modernising what matters first, reducing risk, and improving commercial outcomes without rewriting everything underneath. Read the full customer success story here: https://lnkd.in/gqJmBaeH

Backend modernisation often fails because the cost of getting it wrong is too high. Especially in a large .NET serverless backend on AWS Lambda, where one change can have a wider impact than expected. So progress slows. Or stops. Part 2 of this series, by DiUS Senior Software Engineer Ujjavala Singh, shows what it looks like to keep moving, without betting everything on a rewrite. The challenge isn’t just where to start. It’s how to continue safely. That means: 🔍 understanding how the system behaves before changing it 🧭 making targeted changes where the risk is controlled ♻️ evolving the architecture while delivery continues This is where generative AI helps again, not as a shortcut, but as a way to reduce uncertainty as the system changes. If your backend feels too critical to touch, this is a more realistic path forward. Link in comments.

We’re using generative AI across the SDLC every day at DiUS It’s exciting to experience the step change in how quickly tasks can be completed, but also challenging given the pace of change. That’s why it’s valuable to compare notes with people we trust. Today we hosted long-time DiUS friend and collaborator Andy Kelk for a lunchtime brown bag on how AI is changing the way teams work. Andy had put a LinkedIn post out offering his time to share what he’s seeing. A generous move, and one we jumped at. As this space shifts from AI-augmented engineering toward AI-driven systems where code becomes abstracted away entirely, it’s never been easier to produce code, but much harder to understand, review and trust it. As Andy described, you’re starting to see queues fill up with AI-written code that passes tests, looks right, but where no one can quite explain why it works. This changes how teams deliver, how quality is maintained, and where risk sits. What also came through was what holds things together when this happens. Not more tools, but stronger fundamentals: ✅ Clear intent and better-defined problems ✅ Architecture that provides direction and guardrails ✅ Discipline around quality, testing, and standards These are the same things we’re seeing deliver results in practice. Thanks again Andy for the perspective and the conversation that followed.

Frontend modernisation usually stalls for one reason: No one can agree where it’s safe to start. Especially when you’re dealing with a large, legacy Angular application. The codebase is too big. The dependencies are unclear. And the risk of breaking something in production feels high. So teams wait. Or they plan a rewrite. In Part 1 of this series, DiUS Senior Software Engineer Ujjavala Singh shows a more practical approach: Start by making the system visible using generative AI. Not to generate code. But to: 🔍 map dependencies across a legacy Angular app 🧭 identify safe seams for incremental change ⚠️ surface risk earlier, before it shows up in production From there, you can start replacing parts of the system step by step, without stopping delivery. If 'where do we even start sounds familiar, this is a good place to begin. Link in comments.