Tech talks

What made this talk stand out was the focus on Azure Bicep. I'd been working with Terraform at WhiteSky and considered it the default IaC tool, but they made a strong case for Bicep as Azure's native domain-specific language. The live demos showed how Bicep integrates more tightly with Azure Resource Manager than third-party tools, and how you can build modular, reusable templates that scale across teams. Seeing the progression from ad-hoc scripts to a fully governed deployment pipeline was eye-opening.

What stuck with me most was the framing. I'd always thought of IaC as a technical skill you either use or don't. This talk reframed it as a maturity journey for the whole team, not just one person's workflow. The higher levels aren't just about better tools; they're about governance, scalability, and policy enforcement. That shift in perspective is something I've already started applying, thinking about where our infrastructure practices sit on the scale and what it would take to move up a level.

In our cloud modules at Howest, we covered Terraform basics and deployed some resources, but mostly at level two or three. I think the curriculum could push further into governance and policy-as-code, because that's what companies are actually looking for now. Either way, I'd definitely attend a talk like this again, and I'd drag along anyone who still thinks cloud infrastructure means clicking buttons in a portal.

The live demos were the highlight for me. He showed DeepSeek's reasoning capabilities side by side with proprietary models, tested its code generation, and demonstrated its multimodal features. I went in expecting it to be noticeably worse than the big commercial models, but honestly, the gap was smaller than I thought. For some tasks, especially code and reasoning, it was holding its own, and it's fully open-source. That combination is hard to ignore.

What I didn't expect was how much time we spent on the geopolitical and ethical side. DeepSeek is built in China, and that raises real questions about transparency, data governance, and trust. The speaker didn't shy away from that; he laid out the concerns honestly and let us draw our own conclusions. For my work at WhiteSky, where I was self-hosting AI models through GPUStack, this mattered a lot. If you're going to run an open-source model on your own infrastructure, you need to understand where it came from and what tradeoffs you're making.

We covered ML basics in our studies, but the rapid evolution of open-source AI, and the geopolitical dynamics behind it, wasn't part of the curriculum. I think sessions like this are essential because the landscape changes every few months. I'd recommend this talk to anyone working with AI, even if you're already using commercial APIs. Understanding what open-source alternatives can do, and the context around them, is becoming non-negotiable.

That 2016 decision was the key moment of the talk for me. Recognizing cyberspace as an operational domain fundamentally changed how a 32-nation military alliance approaches digital threats. It meant cyber defence wasn't just an IT problem anymore; it became a strategic priority on the same level as conventional military operations. Hearing that story from someone who was actually at NATO HQ when these decisions were being made was completely different from reading about it in a textbook.

For me personally, this talk added a layer of context to the security work I do. It's easy to get tunnel vision: patch this CVE, harden that config, set up monitoring. But stepping back and seeing how nation-states operate, how critical infrastructure gets targeted, and how international cooperation in cyber defence actually works... it changes how you think about the systems you're building. You're not just protecting a web app; you're part of a much bigger picture.

In our security courses at Howest, we mostly focus on the technical side: network security, ethical hacking, some forensics. The geopolitical dimension is pretty much absent, and I think that's a missed opportunity. Even one session on cyber warfare and how international alliances coordinate their digital defense would give students much better context. I'd recommend this talk to any IT student, not just the security crowd. You don't need to be a pentester to benefit from understanding how the world's biggest alliance thinks about digital threats.

It was a data-heavy, technical session, which I liked. Instead of just saying "IPv6 is important," the speaker showed actual measurement data from Google, Cloudflare, and national statistics. Belgium's numbers were better than I expected, but the global picture made it clear that the transition is far from done. The comparison of different migration approaches (their complexity, security risks, and operational overhead) was practical and immediately useful.

The security angle caught my attention the most. I hadn't really thought about how IPv6 tunneling can create blind spots in network monitoring. If you've got Teredo or 6to4 tunnels running and your firewall isn't inspecting that traffic, you've got an invisible bypass. That's the kind of thing that's easy to miss in a production environment, and it's directly relevant to the network infrastructure work I was doing at WhiteSky with OPNsense and WireGuard.

In our networking courses, we learned IPv6 addressing and subnetting, but the practical migration side was mostly theory. There's a big gap between knowing how to write an IPv6 address and knowing how to actually migrate a production network without breaking things. I'd love to see a lab exercise on setting up dual-stack environments at Howest. This talk convinced me to stop treating IPv6 as an afterthought in my own projects; it needs to be planned from the start.

I'm not a mobile developer and I don't plan to become one, so I wasn't sure how much I'd get out of this. But what made the talk click for me was seeing how Flutter's approach applies to a broader principle I care about: using software to replace expensive, single-purpose solutions. Whether it's a dedicated mobile app for each platform or a manual cloud provisioning workflow, the idea of building once and deploying everywhere resonates with how I think about infrastructure.

The technical side was well-paced: not too deep for newcomers, but with enough real-world case studies to keep it practical. The speaker showed actual apps built with Flutter and walked through the development experience, including the hot reload feature that lets you see changes instantly. That kind of rapid iteration is something I appreciate from my own work, even if I'm working on backend systems instead of mobile UIs.

I'd recommend this talk to students who haven't picked a specialization yet. Seeing how a single framework can target mobile, web, and desktop from one codebase is motivating, and it shows that modern development is about picking the right combination of tools for the job. Would I attend a follow-up? Yeah, especially if it covered Flutter's backend integration or performance at scale.

What I appreciated most was how practical it was. They didn't just talk about models in the abstract; they showed how to find, evaluate, and deploy them using Hugging Face. They also got into the licensing side, which matters a lot if you're thinking about commercial use. Not every "open" model is actually open for business, and understanding the difference between Apache 2.0, Llama's community license, and other terms is something most students don't think about until they have to.

This talk directly influenced decisions I made at work. At WhiteSky, I was building AI automation pipelines with N8N and self-hosted models through GPUStack. The big question was always: do we pay for API calls to a proprietary model, or do we run our own? The ML6 presentation made it clear that for anything involving sensitive data that shouldn't leave your infrastructure, open-source is the only real option. That's not an ideology thing; it's a practical data governance decision.

Our studies covered ML fundamentals, but the tooling ecosystem was mostly a blank spot in the curriculum. I think a hands-on lab where students deploy and compare open models would be incredibly valuable; it's what the industry expects from anyone touching AI infrastructure now. I'd definitely attend another ML6 session, and I'd recommend this one to anyone who works with data or AI, even if you're not building models yourself. Understanding this ecosystem is becoming table stakes.

Before this talk, I understood security mostly from the defensive side: set up firewalls, harden configs, patch things, monitor logs. That's important, but it's reactive. CTI flipped my perspective. Instead of waiting for an alert, you study how attackers think, what they target, and how they chain vulnerabilities together. It's like learning chess by studying your opponent's openings instead of just reacting to each move.

What made it especially valuable was hearing from people who do this professionally at a national level. The practical examples they shared (how real threat campaigns are tracked, how intelligence gets turned into actionable defense) were exactly the kind of insights you can't get from a textbook. It bridged the gap between theory and what actually happens when a threat actor targets an organization.

This session is what pushed me to start building my own offensive security tools, including the NmapToolKit project. I'm also digging deeper into MITRE ATT&CK mapping and working with monitoring tools like ELK and Wazuh. In our courses at Howest, we cover ethical hacking and network security, but the CTI discipline was new to me. I think adding even a single module on threat intelligence platforms and adversary emulation would make the security curriculum significantly stronger. I'd recommend this talk to anyone in IT; understanding how threats actually work makes you better at every aspect of the job, not just security.