CIMSOLUTIONS is looking for a Cloud Architect.
About the company. CIMSOLUTIONS is a behemoth of Dutch IT consulting. They have been recognized as a “Top Employer” 15 times. Their portfolio is filled with serious governmental and quasi-governmental clients: the tax service, the Ministry of Education, the railway operator ProRail. This is a world where the cost of an error is measured not in lost profit, but in a nationwide system failure. Stability, security, and predictability here are not just empty words, but the foundation of everything.
About their current situation, what problem and/or pain they are trying to solve. The job description screams this between the lines. They need not just a tech expert, but a diplomat, strategist, and visionary all rolled into one. A person who can enter a cumbersome government structure and explain in business terms why a hybrid cloud is better than a private one, and why microservices are not just a buzzword. They must take on the full complexity: from choosing between IaaS and PaaS to writing Terraform scripts, all while adhering to strict NIST standards. The company’s pain point is the search for a universal soldier capable of being a translator from bureaucrat-speak to developer-speak, an engineer, and a strategist. Such a person is expensive (up to 8000 euros per month plus bonuses and a car), hard to find, and becomes a bottleneck in any project.
Briefly, how the problem could be solved with AI. What if, instead of searching for this mythical unicorn, they invested in a system that would take on 80% of their routine, yet most critical work? We’re talking about creating or utilizing an AI assistant for the architect. This system could generate optimal, secure, and cost-effective cloud architectures based on business requirements and constraints, freeing up human specialists for that very diplomacy and strategic vision.
Description of approaches, specific tools, implementation steps, ways to reduce distrust.
Let’s imagine a dialogue. Two people are sitting in the meeting room: Dirk, a department head at CIMSOLUTIONS, and I.
Dirk: “We need someone with experience. Someone who has seen ten different migrations to Azure and knows all the pitfalls. Someone who can defend their solution before the board of directors.”
Me: “Absolutely. Now imagine that this person has an assistant who has seen not ten, but ten thousand migrations. Who has analyzed all the documentation from AWS, Azure, and Google Cloud, all NIST standards, all your company’s successful and failed cases over the past 5 years. This assistant doesn’t get tired, has no bias towards a favorite vendor, and is ready to generate three architecture options in 30 seconds with full cost calculation, risk analysis, and ready-to-use IaC code.”
Dirk looks on with distrust. And wrongly so. Here’s how it could work:
Step 1: Creating the system’s “brain.” We take a large language model (LLM), for example, a fine-tuned version of Llama 3 or GPT-4, and “feed” it a closed loop of data:
CIMSOLUTIONS’ internal best practices and architecture templates.
All official documentation and well-architected frameworks from AWS, Azure, GCP.
Security and compliance standards, primarily NIST, but also GDPR and other relevant ones for their clients.
Anonymized data from past projects: what decisions were made, how much they cost, what problems were encountered.
Step 2: Developing the “business requirements -> architecture” interface. Instead of the architect spending hours listening to the client and then drawing diagrams, a business analyst fills out a structured questionnaire. “What is the expected load? What are the fault tolerance requirements? Budget? Data sovereignty requirements?” This data is then transformed into a detailed prompt for our AI assistant.
Step 3: Generation and selection. The AI assistant provides not one, but several solution options. For example:
Option A: Maximally reliable and secure, based on Private Cloud, but 30% more expensive.
Option B: Hybrid model with an optimal price/quality ratio, using PaaS services to accelerate development.
Option C: Fully public cloud with a focus on serverless and maximum cost savings.
Each option includes: an architectural diagram, a detailed 1-year cost estimate, a list of potential risks, a NIST compliance map, and ready-to-use Terraform/Pulumi code.
How to reduce distrust? Start small. Launch a pilot project. Take an already completed successful project and “run” its requirements through the AI assistant. Compare the result proposed by the AI with what a human did. I am confident that in many cases, the AI will suggest a more elegant or cheaper solution. The role of the human architect shifts from “creator” to “curator” and “validator.” They no longer spend days on routine tasks but use their experience to select the best AI-proposed option and adapt it to the client’s nuances.
Description of how to validate AI results.
Trust, but verify. The output of the AI architect should not be taken at face value. The validation process must be rigorous and multi-layered.
1. Expert review. The human, the very Senior Architect they are looking for, reviews the proposed AI architectures. But they do so not from a blank slate, but with several well-developed options before them. Their task is to find logical flaws, account for unformalizable political aspects within the client’s organization, and choose the optimal path.
2. Automated code review. The generated Infrastructure as Code (IaC) is automatically run through static analyzers like Checkov or Terrascan for vulnerabilities and non-compliance with security policies.
3. Sandbox. The best architectural option is deployed in an isolated test environment. Load tests, fault tolerance tests, and (crucially!) simulated attacks by “white hat” hackers are performed on it.
4. Financial audit. The estimate proposed by the AI is cross-referenced with actual cloud provider calculators to rule out “hallucinations” in pricing.
Ultimately, CIMSOLUTIONS gets more than just a replacement for one expensive specialist. They gain a scalable system capable of serving dozens of projects simultaneously, reducing the risks of human error and accelerating time-to-market for their conservative yet very demanding clients. And the very architect they are looking for becomes not a “workhorse,” but the conductor of a high-tech orchestra, which, you’ll agree, is a much more interesting role. And paying 8000 euros a month for a conductor, not a violinist – that’s a whole different conversation.
Источник: https://www.linkedin.com/jobs/view/4415617279/
