“From Geek to Star” #6 - Which Tech in a Fragmented World?

If you missed the previous episodes, you can access them online here.

🗓️ This Week – Chapter 4, Part 2: insights for your work in today's fragmented world

If you remember last week's newsletter (if not, you can read it here before continuing this one), I presented a few global perspectives currently unfolding and asked how this may impact our work as engineers. In this week's newsletter, I will share with you some thoughts that you may take with you and use at work as you discuss with your peers, your leadership, your colleagues of other functions. This will participate to position you as a thinker in your organization.  

🌎🌍🌏 From One Global world to Fragmented Regions

These are the questions I raised in the last newsletter given the current geopolitical shifts. From a tech stack perspective, if you are a US company you may still go fully for US tech stack. If you are a Chinese company, you could go for a fully Chinese tech stack. But what about a US company operating in China or vice-versa? Which tech stack to choose there? And if you are a European, a South East Asian, a Middle East, etc… company? Which tech stacks to use now in which region?

My take on this: designing systems is going to become more complex as we must start to think about more fragmented systems able to take into account the geopolitical impacts. As engineers, we tend to push for more standardization and consistency and rightly so, as this allows for more efficiency. However, we cannot consider building a single system anymore which can cover the whole world (in case you are working in a multinational corporation). Architecting for a complex world is going to become a highly rewarded skill. These are some principles I would be looking at: 

• Leveraging as much as possible on open source solutions and on AI to build and operate core systems. Open source solutions give the benefit of being able to be less dependent on vendors, at a time when the nationality of vendors may become a potential threat or risk for companies. With AI, this may allow for strong internal teams to contribute and add features to open source solutions to improve them even more (and feed back to the community). 

• One corollary of the previous principles is the need to have strong internal engineers. Indeed, this will mean being able to manage more technical challenges than using vendors' products. For example, if using Docker Swarm or Kubernetes, that will require stronger skills than if just using cloud provider's managed services such as EC2. But again, I think that AI will enable good engineers to uplift tremendously, provided companies realise this and value them accordingly. 

• By doing so, that will open the possibility for companies to be less dependent on specific vendors and keep the possibility to move if some challenging geopolitical shifts happen. For example, for a European company, it could decide in this case to put part of its stack on OVH cloud for its customers in Europe if full exposure on AWS, Azure, GCP, Alicloud… becomes too risky, while still being on one of these cloud providers outside if OVH cloud is not performing enough outside Europe. 

Of course, transitioning to open-source stacks can be empowering but requires upfront investment in skills, tooling, and internal support. Digital native companies are probably already well ahead in this game as they embed Tech in their core. For more traditional companies, I would advocate to set up first a small team of internal senior and forward looking people who can start to build and be fully augmented with AI, and then capitalize on this as this team could start to mentor and inspire more teams in a progressive rollout as replacing legacies is always a difficult organizational exercise. 

🤐 No more travel for Sensitive Data?

On this aspect, I raised these questions last week: what does it mean regarding how we architect our systems? With the advent of the cloud, we did not have to care about where the data were stored, just looking at data replication across regions for the sake of performance if the market to address was global. But with these growing regulations on data, are we losing the benefits of the cloud?

In terms of application and data architecture, I believe we will need to include the notion of geography in our data as we build systems in the future. Typically, customers’ data may need to be physically localized in different countries: applications should therefore be more distributed, with for example a Data API layer which will integrate compliance and regulations to process the data and store them accordingly. There should be in the teams some engineers who are proficient in data compliance and regulations to design APIs and data pipelines / storages accordingly, while abstracting the complexity for the other layers with the proper anonymization. 

Leveraging cloud-native architectures like data mesh or deploying regional data lakes with localized access controls can help balance compliance with performance. Designing for policy-aware processing becomes a necessity in today's world. 

💲 The Tariffs war

The point I raised last week was that the tariff war could lead into a rise in the price of the technologies and products used. If you are a European company, should you re-think about your tech stack? What would be the alternatives? 

On this point, I believe that a larger use of open source and AI to augment engineers could help companies to reduce their dependencies and thus exposure to the risk of price increase of the tariffs on digital technologies. 

🧠 Preserving cultures, languages, regional specificities in an AI world

The point I raised last week was about how AI is not neutral from a language, content, or cultural perspective. When designing systems for different countries, will you therefore make it simple at the risk of oversimplification or even distorted outcomes or will you integrate the different specificities at the risk of complexity of your systems?  

I myself do not have a clear idea on how this may be architected. But as we are entering in a world of agentic AI where MCP (Model Context Protocol), when ready for production, will enable agents to connect much more seamlessly to each other and to systems, I think that it is possible to imagine a design where we may have different "knowledge / cultural agents”, able to interact according to the customers. 

Let's take an example in travel: we may have an “American AI agent” able to converse with American customers built on OpenAI or Anthropic LLMs, “an European AI agent” able to converse with European customers built on Mistral LLM,  “a South Asia agent" able to converse with South East Asian customers built on SEA Lion LLM and “a Chinese AI agent” able to converse with mainland Chinese customers built on Deepseek. There would probably be a common layer of business context feeding the different LLMs with the specificities of the company, but the way you address, you interact with customers would respect cultural nuances rather than imposing a main culture to all.  

Ensuring seamless collaboration between culturally tailored AI agents and existing platforms will require standards like MCP and robust API governance (companies should really push for API-first policy, as what Jeff Bezos required for Amazon more than 20 years ago). Inter-agent orchestration will become as critical as agent design itself.

🙏 I’d Love to Hear From You:

These are just a few thoughts on how the current global context may be taken in our work. And how you could use these perspectives to open conversations around you at work, thus starting to position you among people who are able to go beyond pure technical conversations. And slowly moving the needle in terms of the perceived value you can bring, in a time where clarity is much needed on how tech can change the world. 

If you have thoughts on the above and how we could engineer for this fragmented world, do share them with me! 

And feel free to follow me on LinkedIn for short-form reflections and updates between newsletters.

✨ Stay curious, stay connected!

From Geek to Star by Khang | The Way Forward