Blerina Kapllani

Excited to see the launch of something I believe is a genuinely important step forward for infrastructure at scale. Today HashiCorp announced HCP Terraform powered by Infragraph (public preview) and this is more than just an incremental feature release. At its core, this is about solving one of the hardest problems in modern cloud environments: understanding infrastructure as a system, not just as code. Most organisations have invested heavily in Infrastructure as Code, but visibility still lags behind. Teams can provision quickly, but answering simple questions remains hard: - What depends on this resource? - What will break if I change it? - Where are the risks in my environment? Infragraph changes that. Looking forward to seeing how customers start to use this, especially those operating at scale across multi-cloud and hybrid environments. If you're working on platform engineering, cloud governance, or large-scale Terraform estates, this is worth a close look. #Terraform #PlatformEngineering #Cloud #DevOps #InfrastructureAsCode #HashiCorp #Innovation https://lnkd.in/gF324u3G

18

Great interview with Martin Fowler on the Pragmatic Engineer podcast: https://lnkd.in/eFGGbNMG . Martin shares his view on embedding AI in business logic, how the non-deterministic nature of AI affects how software engineers need to approach problems, and how this shift is comparable to moving from writing assembly code to higher level programming languages.

10

https://lnkd.in/dKKyEjFa In the rapidly evolving age of generative AI, senior leaders must focus on leadership and organizational transformation rather than solely on the technology itself. To succeed, they should cultivate AI fluency through diverse networks and cross-industry conversations, redesign organizational structures to maximize AI's value, facilitate collaborative decision-making between individuals and AI, empower teams with coaching and psychological safety, and lead by example in experimenting with AI to encourage widespread adoption. These five critical skills will enable leaders to steer their organizations through the necessary changes to fully leverage AI's capabilities. #AI #LeadershipSkills

11

1 Comment

this is an interesting take and there are some pretty impactful things that claude skills can do, but they run isolated with no outside network access, so they serve a different purpose than mcp does. yes skills are more elegant, but also can't do the same things as MCP. API & MCP serve different purposes from each other as well. I think MCP serves massive value (but needs governance via a catalog & mcp gateway) Yes try out skills and figure out which logic belongs where!

5

Reliability and recovery metrics measure the resilience of your delivery ecosystem and your organisation’s ability to respond under pressure. These metrics answer a critical question: When something goes wrong, how fast can we detect it, contain it, and restore service?

1

How do we find scalable ways to build, impactful ways to lead and robust models to secure.   That was my recent focus, to reset, and refresh with a set of learning courses and certifications to challenge existing thinking, uplift existing certifications and ensure I can support our teams in the best way. I chose the x3 courses below over Oct/Nov (now all completed) to deliver that:   - AWS Certified Solutions Architect - DevOps Institute Certified DevOps Leader - ISC2 Certified Information Systems Security Professional (CISSP) I'd love your thoughts on what i should pick up next year to continue this journey, which course have you recently completed thats left a positive impression / impact? Key reminders / takeaways from each:   ☁️ AWS Certified Solutions Architect 🔑 Design for Resilience, Not Recovery. Using cloud-native approaches (like the AWS Well Architected framework) to building fault-tolerant systems from day one looking at what will happen 'when' not 'if' a fault occurs - as we're doing inside our engineering teams at Natwest 🔑 Cost & Operational Control is a Design Pillar. Optimizing for cost, security and operational delivery is a continuous architectural duty. It means selecting the right service (a lot of emphasis in the course on right-sizing service choices) eliminating undifferentiated heavy lifting.   👥 DevOps Institute DevOps Leader 🔑 Culture is the biggest Enabler. It’s a shift built on psychological safety, shared goals, and breaking down organizational silos. Some great reminders in this course on stakeholder and team engagement models, different organisational culture norms and levers to help shift 🔑 Focus on Flow, Not Just Speed. The goal isn't just to be fast but to look holistically at the value stream to eliminate waste, reduce handoffs and challenge queue time. Something we've taken to heart in our Route to Live simplification engineering work.   🛡️ ISC2 Certified Information Systems Security Professional 🔑 Security is everyone's priority. A useful reminder of the wider aspects of security from a purely technical software discipline to a core business risk management & assessment - with timely reminders of the causes / impact of Stuxnet, Solarwinds and similar prior attacks. 🔑 There is best practice out there - use it. Some useful deep-dives into security frameworks, threat modelling, standards and practices - covering NIST, ISO, SOC, OWASP (and many more) along with useful refreshers on core cert, cryptographic and physical security & risk standards.   For those interested, I would highly recommend them, all able to be completed online in a self-paced manner using some helpful online training materials to get into the detail of the core concepts.   #AWS #DevOps #CyberSecurity #CloudComputing #Leadership #TechSkills

167

9 Comments

APIs in 2025 = AI Strategy Postman's latest report (5,700+ devs): 📈 82% API-first (+12% YoY) 💰 65% generate revenue from APIs 🤖 89% use AI daily ⚠️ Only 24% design for AI agents 🔒 51% fear unauthorized agent access Translation: We're building with AI but not for AI. Gap = Opportunity (or crisis, depending on speed). Machine-readable schemas, agent-ready security, and product thinking around APIs aren't nice-to-haves anymore. They're table stakes. Full report 👇 https://lnkd.in/ed7qgsaf #APIs #AI #TechStrategy #DevTools

47

1 Comment

🚀 UK DevOps & SRE Engineers — This Changes How You Ship Code CI/CD failures, late-night alerts, broken deployments — they’re not part of the job, they’re automation gaps. We’re introducing a FREE DevOps Automation Trial designed for UK-based engineering teams who want: ✅ Faster & reliable deployments ✅ Automated CI/CD pipelines ✅ Docker & Kubernetes readiness ✅ Early error detection ✅ Instant alerts before customers notice ✅ Less firefighting, more engineering 🤖 Powered by AI Automation, this solution doesn’t just deploy — it observes, predicts, and fixes issues before they escalate. This is not another tool. This is DevOps + SRE intelligence working together. If you’re serious about: • System reliability • Release velocity • Production stability • Scaling without chaos 💬 DM me to start your FREE DevOps Automation Trial Let automation handle failures — you focus on building the future. #DevOps #SRE #UKTech #CloudEngineering #CICD #Automation #AIinTech #PlatformEngineering #Kubernetes #Docker #CloudNative #TechLeadership #EngineeringExcellence #StartupUK #ScaleWithAI

4

🧠🤖 LLM vs SLM: What’s the Difference and Why It Matters As AI adoption grows, we’re hearing two terms more often: 👉 LLM – Large Language Model 👉 SLM – Small Language Model But what’s the difference? And when should you use one over the other? Let’s break it down 👇 🔹 Large Language Models (LLMs) Think: GPT-4, Claude, Gemini, Mistral 7B+ ✅ Trained on massive datasets ✅ Billions of parameters ✅ Capable of deep reasoning, multi-turn conversation, coding, summarization ✅ General-purpose, high accuracy, high context window But also: ⚠️ Expensive to run ⚠️ Slower inference time ⚠️ Requires more memory & compute Use when: You need strong general intelligence The task involves open-ended reasoning or creativity Latency and cost are not blockers 🔸 Small Language Models (SLMs) Think: Phi-2, TinyLlama, DistilBERT, Falcon-R, Gemma 2B ✅ Lightweight and fast ✅ Cost-effective ✅ Ideal for edge devices or tightly scoped enterprise use ✅ Easier to fine-tune and deploy privately But: ⚠️ Narrower capabilities ⚠️ Limited context length ⚠️ May struggle with complex reasoning Use when: You have specific tasks (e.g., intent detection, classification, simple Q&A) You care about speed, cost, or privacy You want to deploy on edge or mobile 🚀 Real-World Analogy LLM = Swiss Army knife — powerful, can do almost anything SLM = Precision tool — lightweight, perfect for one job In the future, most orgs will run: ✅ LLMs in the cloud for heavy lifting ✅ SLMs at the edge or in-app for fast, focused tasks Know the difference. Choose the right tool. Build smart, scalable AI. #LLMvsSLM #AIEngineering #SmallLanguageModels #LargeLanguageModels #AIArchitecture #EdgeAI #TechLeadership

20

8 Comments

Found a 💎 while working on this week’s weekly digest Scalable Enterprise Java for the Cloud is a free eBook published by Payara in collaboration with Java ecosystem experts. It’s aimed at helping developers understand how to build modern, cloud‑native enterprise Java applications using Jakarta EE as the core standard. https://lnkd.in/gDSkSvhV

27

My RCA CDP1802 toolchain took another step today when it managed to build gcc's libstdc++ for the target. Not the scaled down freestanding one, but the full-on hosted one. The target is an 8-bit micro with 16-bit address space, so it took a bit of tweaking to find the right set of options to make the code fit. Eventually the following gcc configuration options worked: --enable-target-optspace --enable-cxx-flags=-g0 --disable-libstdcxx-verbose --disable-libstdcxx-backtrace The initial C++ test suite run showed a 2.5% failure rate. Many failures are due to code size or a conflicting declaration with my libc, but I haven't started debugging them. In comparison, the C test suite has a 0.2% failure rate. The libstdc++ test suite doesn't run at all due to code size of some testsuite_abi.cc it wants to compile first.

8

I've just published a new article about transforming deployment culture, in particular how we went from releasing every 6 weeks to 5.1 times per day, and why the solution wasn't revolutionary technology, but "Vanilla Scrum." The barriers weren't technical. We had CI/CD and feature flags. The real blockers were organisational: lengthy approval processes, adversarial department relationships, and defensive workflows. The breakthrough? Basic Agile fundamentals done well: → Small talented teams → 2-week sprints with proper estimation → Collaborative refinement over lengthy specs → Trust-based processes I call it Vanilla Scrum because there's nothing wrong with doing the basics well. It's often exactly what organisations need. Read the full transformation story here: https://lnkd.in/eG_Bx7aW The natural state of software development is frequent releases. Everything else is organisational friction. What's preventing your teams from deploying frequently? #DevOps #TechLeadership #AgileTransformation #VanillaScrum

8

What makes software systems work beyond features? Building the RIGHT thing isn’t enough. Building the thing RIGHT isn’t enough either. Great software systems don’t succeed on features alone . They succeed (or fail) based on how well we design and balance the multiple dimensions that shape a system: business value, security, reliability, performance, cost, operability, maintainability, interoperability, and usability. In this blog, What makes software systems work beyond features (https://lnkd.in/e-pwdD4i), I explore software systems as multi-dimensional constructs where excellence only emerges when we build the RIGHT thing, the RIGHT way. 👉 Which dimension do you see most commonly overlooked or under-invested in? I’d like to hear your view

17

1 Comment

We used to move money synchronously. It worked - until it didn't. At Equals Money, we replaced legacy synchronous flows with a modern event-driven platform that now processes billions in cross-border payments. That shift didn't happen overnight, and it didn't happen without hard lessons. Next week at AWS Summit London, I'm joining Panayiotis-Orestis Antoniou and Amir Khairalomoum to talk through what that architecture journey looks like - from event source mappings with SQS and Kinesis, to orchestration with Lambda Functions and routing with EventBridge. We'll cover the real patterns behind production event-driven systems: where they shine, where they bite you, especially around error visibility and distributed tracing, and how to think about the trade-offs before you're deep in a migration. Come along if you want practical takeaways you can bring back to your team. 🎤 #AWSSummitLondon #AWSSummit #AWSLambda #Serverless #EventDriven #FinTech

24

Consistency in distributed systems isn’t binary. It’s a business trade-off. Immediate consistency feels comforting, every read reflects the latest write. But at scale, this comes with trade-offs like latency and tight coupling. Eventual consistency embraces reality: writes propagate asynchronously, systems stay available under failure, and resilience improves. But it comes with trade-offs like potentially stale data. The real question for architects isn’t “Which is better?” It’s: - What’s the business tolerance for staleness? - Where do we need strong guarantees (e.g. balances)? - Where can we accept lag (e.g. recommendations, analytics)? - Modern systems blend both. It’s not a purity contest, it’s a design choice tied to business risk. What’s the toughest consistency trade-off you’ve had to make in a real system? #EventualConsistency #SystemDesign #ArchitectureDecisions #DistributedSystems #EDA #DDD

22

1 Comment

Over the last few years, many have turned to Cloud Composer (Apache Airflow) for orchestration on GCP. However, as platforms evolve toward cloud-native, serverless, and API-driven architectures, it’s essential to consider: Are we still using Airflow out of necessity or simply because it’s familiar? This post outlines a practical strategy for moving beyond Composer by leveraging native GCP services like Workflows and Cloud Run Jobs. The focus is on simplicity, cost efficiency, and operational clarity. Insights on how others are modernizing orchestration on GCP would be appreciated. #DataEngineering #GoogleCloud #ApacheAirflow #Serverless #CloudArchitecture

30