Mohamed Allam

Last month, at ADC, I gave a talk about something that sits at the heart of modern software development: teams. In software engineering, we still celebrate the myth of the hero: the individual who ships fast, fixes everything, and saves the day. Heroics can feel efficient in the moment, but they make organisations fragile. They don’t scale, they hide systemic problems, and they leave teams exposed when knowledge lives in one place. The core idea of the talk was simple: Great software isn’t built by great individuals. It’s built by great teams, and great teams are designed, not accidental. I spoke about: - Why reliance on individual heroics creates brittleness, technical debt, and poor long-term outcomes - What research tells us about high-functioning teams: stable membership, clear purpose, sound structure, the right mix of skills, supportive environments, and space for reflection - Why many hardware-first organisations struggle as software becomes central to the product experience - How collaboration breaks down when teams are optimised around functions instead of outcomes I also shared how we’re experimenting at Focusrite Novation to design better conditions for collaboration: - Stable, outcome-oriented teams that stay together long enough to build trust and accountability - Shared planning across engineering, product, and design, focused on why before what - Making dependencies, flow, and progress visible across disciplines - Lightweight reflection loops that prioritise learning over ceremony We’re still on the journey, and this isn’t a solved problem. But the shift has been clear: when you design the right conditions, collaboration stops being something you hope for. It becomes the default. The question I left the room with was this: Are we designing for teams, or are we relying on individuals to carry the system? Thanks to everyone who attended, asked thoughtful questions, and shared their own experiences. And thanks to the ADC organisers for creating space for these conversations.

37

1 Comment

🚀 Two days before WhatTheStack kicks off, I’m running a crash course in building apps that survive the real world. Real-World React: The Architectural Playbook for Performance, Resilience & Scalability is built around React and Next.js as our shared language, but it’s not exclusive. I’ve had Vue, Svelte, and Angular devs attend, and most of the concepts we’ll cover are framework-agnostic and apply to any modern frontend stack. We’ll dig into: - Architecture patterns that implicitly boost performance - The real story behind React reconciliation (and anti-patterns to avoid) - Resilience engineering for modern frontends - DORA metrics and delivery velocity at scale - Building post-deployment confidence so you can ship without fear This isn’t about pixel-perfect UIs, it’s about patterns, decisions, and trade-offs that keep apps fast, resilient, and scalable when they’re serving millions, not just running locally. 🎯 If you work in frontend and want to level up your architecture game, this is for you.

5

DevOps engineers are becoming the most strategic hires in Germany. But not for the reasons many think. It’s no longer just about: • CI/CD pipelines • Kubernetes • Infrastructure as Code The new shift is AI-enabled DevOps. Companies are starting to integrate: ⚙️ AI-assisted monitoring ⚙️ Predictive infrastructure scaling ⚙️ Automated incident response ⚙️ AI-generated code reviews This means the modern DevOps engineer needs to understand: • Cloud platforms • Automation • Observability • AI tooling And here is the challenge: The number of engineers who truly combine DevOps + AI thinking is still extremely small. This makes them one of the most valuable profiles in today's IT market. Hiring managers: Are you already seeing this shift in your DevOps teams? #DevOps #PlatformEngineering #CloudEngineering #AIinTech #Kubernetes #TechHiring #GermanyIT

13

𝗧𝗵𝗲 𝗯𝗲𝘀𝘁 𝗴𝗶𝗳𝘁 𝘆𝗼𝘂 𝗰𝗮𝗻 𝗴𝗶𝘃𝗲 𝗮 𝘀𝗽𝗲𝗮𝗸𝗲𝗿 𝗶𝘀𝗻’𝘁 𝗮𝗽𝗽𝗹𝗮𝘂𝘀𝗲, 𝗶𝘁’𝘀 𝗳𝗲𝗲𝗱𝗯𝗮𝗰𝗸. After my sessions at the International JavaScript Conference Munich, I got the feedback summary this week. Both talks scored really well: 🟩 𝗖𝗮𝗰𝗵𝗶𝗻𝗴, 𝗣𝗮𝘆𝗹𝗼𝗮𝗱𝘀, 𝗮𝗻𝗱 𝗢𝘁𝗵𝗲𝗿 𝗗𝗮𝗿𝗸 𝗔𝗿𝘁𝘀: 𝗢𝗽𝘁𝗶𝗺𝗶𝘇𝗶𝗻𝗴 𝗨𝗫 𝗶𝗻 𝗦𝘂𝗯𝗼𝗽𝘁𝗶𝗺𝗮𝗹 𝗖𝗼𝗻𝗱𝗶𝘁𝗶𝗼𝗻𝘀 • Quality: 4.67 / 5 • Knowledge: 5 / 5 • Comment: “A bit overloaded with information.” 🟦 𝗙𝗿𝗼𝗺 𝗙𝗿𝗮𝗴𝗶𝗹𝗲 𝘁𝗼 𝗙𝘂𝘁𝘂𝗿𝗲-𝗣𝗿𝗼𝗼𝗳: 𝗦𝗼𝗹𝘃𝗶𝗻𝗴 𝗪𝗲𝗯 𝗠𝗼𝗻𝗲𝘁𝗶𝘇𝗮𝘁𝗶𝗼𝗻 • Quality: 5 / 5 • Knowledge: 5 / 5 It’s always nice to see high numbers; it means the work, preparation, and energy paid off. But the most valuable part was that small comment: “A bit overloaded with information.” I actually love that. It’s true, I tend to pour a lot into my talks because I want people to walk away with something practical. The tradeoff is that sometimes it becomes dense. That single line says more to me than a perfect score ever could. It’s what helps me get better. Feedback like this is 𝘂𝗻𝗱𝗲𝗿𝗿𝗮𝘁𝗲𝗱. Most people don’t realize how little of it speakers actually get after a talk. You stand on stage, share your work, and often walk away without knowing what really landed and what didn’t. When someone takes 20 seconds to write one honest sentence, it gives you a mirror. It shapes your next talk, your style, your pacing, and your focus. So if you’ve ever attended a meetup, workshop, or conference talk, take that moment to leave feedback. Whether it’s praise, confusion, or critique, it all helps. And to everyone who came to my sessions in Munich, thank you. You showed up, listened, asked, challenged, and cared enough to respond. That’s what makes these communities worth being part of. #IJS #PublicSpeaking #Conferences #EngineeringCulture #Feedback #ContinuousImprovement #Learning #DeveloperCommunity

21

3 Comments

Alright, #WeAreDevelopers Berlin. On Thursday, July 10, I will be doing *TWO* talks -------------- 12:10-12:40, MICROSOFT Stage 3 (limited space, come early) Introducing JSON Structure - A Better Schema Mostly everyone uses JSON Schema in some way and mostly everyone had also stepped on their own toes using it. Why? Because we all want a data definition language to define data structure and JSON Schema isn't one. In this talk, you'll learn about JSON Structure, a data definition language that looks quite familiar, but provides better guardrails and fewer pitfalls -------------- 14:10-14:40, Stage 7 Bringing Clarity to Event Streams: Enabling Analytics and AI Through Rich Metadata Event streams are central to delivering actionable insights for analytics and AI. However, without rich metadata to contextualize these streams, data consumers often struggle with inconsistencies, poor quality, and a lack of interpretability. By leveraging metadata as a cornerstone of event stream architecture, organizations can unlock clarity in their streaming data, bridging the gap between raw event streams and actionable insights. This clarity not only enhances the quality and usability of event streams but also ensures alignment between developers, analysts, and AI systems, enabling them to extract maximum value from data in motion. This session explores how rich metadata transforms event streams into a reliable foundation for advanced analytics and AI workflows. We will examine the role of schema definitions, semantic annotations, and cataloging practices in improving data quality and context. #wad #berlin #conference

22

4 Comments

Now that I'm using CC so much, I so got used to TUIs, that I don't really want to leave my terminal anymore! So here is my terminal stack: * lazygit — a nice TUI git client with worktree support * yazi — "Blazing fast terminal file manager written in Rust, based on async I/O" * fresh — simple but powerful terminal text editor * ghostty — fast terminal emulator. I've been using it for more than a year, but the split panel feature really shines when you want to open any of the apps next to a Claude Code session * lynx — a battle-tested web browser 😉 What is your tooling of choice? Links in comments.

8

11 Comments

Most CI/CD pipelines have a silent killer. It isn’t from failed builds. Speed is a priority for most teams but the basics get overlooked. • When did the build times get longer? • Why did the pipeline freeze? No error, no alerts just nothing. • Why did this take 30 minutes a couple of days ago but 5 minutes today? Here’s what seperates good pipelines from great ones. • Parallelise when possible - running independent jobs simultaneously to save time. • Add alerts - a hanging job is worse than a broken one, it saves you time. • Run your smaller checks first - linting, security scans etc. No need to spin up a full environment for something that can be caught in 30 seconds. • Monitor the pipelines flakiness - track the steps that fail inconsistently and fix them instead of hitting re run. A pipeline you can’t trust is like a flip of a coin. Your pipelines is production infrastrucure, treat it like one.

19

1 Comment

Lightstep sunsetting (https://lnkd.in/e-bkP_Yw) clearly defines the end of the era, but it is not the end of the world, here at Zalando we created and published the Lightstep receiver - a component for smooth transition of trace ingestion from Lightstep tracers to OpenTelemetry. Blog -> https://lnkd.in/eyT6G2qu Code -> https://lnkd.in/eVfMUabW

34

𝐓𝐞𝐫𝐫𝐚𝐟𝐨𝐫𝐦 𝐁𝐞𝐬𝐭 𝐏𝐫𝐚𝐜𝐭𝐢𝐜𝐞𝐬 𝐲𝐨𝐮 𝐧𝐞𝐞𝐝 𝐭𝐨 𝐤𝐧𝐨𝐰 👀 1️⃣ 𝐌𝐨𝐝𝐮𝐥𝐚𝐫𝐢𝐭𝐲 – when writing code, make sure you follow the 𝐃𝐑𝐘 𝐩𝐫𝐢𝐧𝐜𝐢𝐩𝐥𝐞 (don’t repeat yourself). Use modules for reusable code, so you don’t waste time setting up the same infrastructure multiple times. 2️⃣ 𝐕𝐚𝐫𝐢𝐚𝐛𝐥𝐞𝐬 – don’t hardcode values directly into your Terraform code. Use variables instead, so you can easily change values in one place without having to search through your files to update them. 3️⃣ 𝐂𝐫𝐞𝐝𝐞𝐧𝐭𝐢𝐚𝐥𝐬 – never hardcode sensitive information like your AWS access key or secret key. Use the 𝐞𝐱𝐩𝐨𝐫𝐭 command to store them in 𝐞𝐧𝐯𝐢𝐫𝐨𝐧𝐦𝐞𝐧𝐭 𝐯𝐚𝐫𝐢𝐚𝐛𝐥𝐞𝐬 instead. This way, they won’t be exposed in the Terraform state file after you run terraform apply. 4️⃣ 𝐅𝐨𝐫𝐦𝐚𝐭𝐭𝐢𝐧𝐠 𝐚𝐧𝐝 𝐕𝐚𝐥𝐢𝐝𝐚𝐭𝐢𝐨𝐧 – run terraform fmt and terraform validate regularly. This keeps your code clean, consistent, and free of basic errors before applying changes. 5️⃣ 𝐍𝐚𝐦𝐢𝐧𝐠 𝐂𝐨𝐧𝐯𝐞𝐧𝐭𝐢𝐨𝐧𝐬 – use clear and consistent naming for resources, variables, and modules. This makes your code easier to read, understand, and maintain over time. 6️⃣ 𝐒𝐭𝐚𝐭𝐞 𝐌𝐚𝐧𝐚𝐠𝐞𝐦𝐞𝐧𝐭 – always use a remote backend (like 𝐒𝟑 with 𝐃𝐲𝐧𝐚𝐦𝐨𝐃𝐁 for 𝐬𝐭𝐚𝐭𝐞 𝐥𝐨𝐜𝐤𝐢𝐧𝐠) to store your Terraform state file. This ensures your state is consistent, secure, and can be shared safely across your team. What other best practices would you include in this list? Leave it in the comments below 💬 #DevOps #Cloud #Terraform #TerraformBestPractices

19

Stop Maintaining "Zombie" Apps. Start Building Agentic Touchpoints. 🧠✨ We are witnessing the biggest shift in software since the cloud: The move from Static Legacy Systems → Agentic Touch Applications. If your application still waits for a user to click a button to do something, it’s already falling behind. Here is the blueprint for the transformation and why it matters for your bottom line. 👇 🔄 The Transformation: From "Legacy" to "Agentic" Think of your current app as a Library (you have to go find the book). Think of an Agentic App as a Concierge (it brings you the book before you ask). The Flow: 1️⃣ Input: User Intent / Raw Data 🗣️ 2️⃣ Agentic Layer: The AI doesn't just "process" — it Reasons, Plans, and Acts. ⚙️ 3️⃣ Output: A completed outcome, not just a search result. ✅ (This isn't just an upgrade; it's giving your software a brain.) 📈 The ROI Impact (The "Big" Shift) Why does this matter for the CFO? 📉 Slash Operational Costs: Agents don't just "assist"; they resolve. Expect 40-50% reduction in Level 1 support costs as agents handle complex workflows autonomously. ⚡ Speed to Market: Modernizing with Agentic AI allows for rapid refactoring. What took months of manual coding now takes weeks. 💎 Customer Retention: A reactive app is annoying. A proactive agent is sticky. Higher retention = Higher LTV (Lifetime Value). 🎨 Creating "Hyper-Personalized" Service How do you turn a standard login into a VIP experience? The Agentic Touch Recipe: Step 1: Context Awareness 🕵️ Old Way: "Hello, User." Agentic Way: "Hello, Sarah. I see your last payment failed. Want me to retry it with the card on file?" Step 2: Proactive Action 🚀 Old Way: User searches for "Upgrade Plan." Agentic Way: Agent detects high usage usage and suggests: "You’re hitting limits. Shall I auto-upgrade you to save $20/month?" Step 3: Memory 🧠 The agent remembers preferences, not just settings. It learns your friction points and smooths them out automatically. 🚀 The Outcome Legacy apps = Digital Debt. 📉 Agentic Touch apps = Digital Assets. 📈 My Question to you: If your core application could do one task completely autonomously for your users, what would you want it to be? Let’s brainstorm in the comments! 👇 #AgenticAI #LegacyModernization #DigitalTransformation #AI #TechTrends

7

𝗙𝗶𝗻𝗶𝘀𝗵𝗲𝗱 𝗺𝘆 𝗽𝗼𝗿𝘁𝗳𝗼𝗹𝗶𝗼 𝗶𝗻 𝟮 𝗱𝗮𝘆𝘀, and I’m calling it done (for now) When people see portfolio sites, they often think it took weeks or months. But this one? I built it in 48 hours. 𝗛𝗼𝘄? 1. I leaned on Loveable AI + ChatGPT to help generate layout ideas, copy, and UI tweaks 2. I researched top portfolios from designers and devs around the world, took inspiration, not copy 3. I stitched it all together with React, TypeScript, TailwindCSS 4. I set up CI/CD and deployed to Azure Static Web Apps with a custom domain Integrated a serverless contact form (Formspree) and scheduling (Calendly) What I wanted was not a portfolio that just looks pretty, but one that works end-to-end: fast, secure, interactive, responsive. And you can do the same, you don’t have to be a senior dev to build something polished now. If you’re building your portfolio this week, here are my tips: 1. Use AI for ideation — blog layouts, mockups, copy — but always revise to your voice 2. Start simple, then layer in polish (animations, background accents, interactions) 3. Use a modern stack you understand deeply (I used React + TypeScript + Tailwind) 4. Automate your deploys (CI/CD) so every push goes live 5. Always include a way people can contact you gracefully (Formspree/Calendly) I’d love to see your portfolios, reply with a link and I’ll check them out, or ask me any questions about how I built mine. Live version: lolaakinrinsola.com Tech stack: React, TypeScript, TailwindCSS, Azure Static Web Apps, Loveable AI, ChatGPT, Formspree, Calendly #portfolio #webdev #frontend #cloud #reactjs #azure #typescript #devlife #ai #buildinpublic #canadatech

17

3 Comments

You do not have a model problem, you have a discovery problem. The answer is not picking the best model. The real answer is asking the right question first. During Week 2 of Building with Agentic AI, one idea became very clear to me: Choosing the right Bedrock model as a Solution Architect is more about design thinking than just model selection. Yes, you can use multiple models in a system and sometimes that is actually the best architecture. But before selecting any model, remember these four pillars: 1️⃣ Capability – Is the model capable of handling the complexity of your task? 2️⃣ Features – Does it provide the tools required to solve your business case? 3️⃣ Latency – Can it respond fast enough for your application? 4️⃣ Cost – Is it cost-effective at scale? Sometimes the smartest architecture is not using a single premium model for everything. Instead, design your workflow like this: Prepare → Compare → Ship Different stages may use different models optimized for that specific task. 💡 Right model per task = optimized cost + maximum accuracy In many real-world architectures, this approach can reduce costs by 60–70% compared to using a single premium model for every step. AI architecture is not about chasing the most powerful model, it’s about designing intelligent systems. Special thanks to Parna Mehta Prasad Rao and Ashish Prajapati for the guidance and inspiration.

5

1 Comment

Why "Caching Plugins" Aren't Enough Anymore Think a caching plugin will fix your slow REHub site? Think again. ❌ Plugins like WP Rocket are great, but they can't fix: Third-party API delays (Content Egg). Unoptimized database queries from heavy themes. Main-thread blocking from bloated JS libraries. In the Banking in Bim migration, we didn't just "cache" the problem—we eliminated it. By moving to a custom Next.js stack, we removed the need for 10+ "optimization" plugins that were actually adding to the bloat. Stop patching. Start architecting. #WebDev #CleanCode #Optimization #WordPressTips #HeadlessDevelopment

2

1 Comment

The Kroxylicious project is very pleased to announce the release of Kroxylicious 0.16.0 [1]. See the Changelog [2] for a list of changes and summary of Deprecations, Changes and Removals. Apache Kafka 4.1.0 compatibility - The main change in this release is Apache Kafka 4.1.0 compatibility. The proxy will now relay Kafka 4.1 specific versions of the Kafka RPC and your filters can intercept them. Observability improvements - Thanks to the work of hrishabhg[3], the proxy now sports connection count gauges, allowing you to monitor downstream and upstream connection counts in realtime. [1] https://lnkd.in/gRw5a_SU [2] https://lnkd.in/gRUJ9h2J [3] https://lnkd.in/gxaPwviu

26

🚀 Why Distroless Images Are Changing the Container Game 🚀 If you’re building containers, you’ve probably heard about Distroless Images — but why are they gaining so much momentum? Unlike traditional container images that bundle a full OS, distroless images strip everything unnecessary out — no package managers, shells, or extra tools. Just your app and its runtime dependencies. Benefits at a glance: ✅ Smaller image size = faster downloads & deployments ✅ Reduced attack surface = better security ✅ Simpler, cleaner builds = fewer vulnerabilities ✅ Ideal for production environments where minimalism and security matter most If you want containers that are lean, secure, and efficient — distroless might be your next go-to strategy. Have you tried distroless images yet? Share your experience or questions below! 👇 #Containers #Kubernetes #DevOps #CloudNative #Distroless #Cybersecurity #SoftwareDevelopment

8

Our Terraform pipeline ran perfectly for months. Then one morning — random failures started appearing. Sometimes deployments succeeded. Other times, Terraform complained: “Resource already exists.” Or worse — state lock errors. At first, we thought it was a cloud API issue. Then we realized — it was our own CI/CD behavior. Terraform is stateful. When multiple pipelines run in parallel, they often compete for the same state file, leading to race conditions and partial updates. Over time, that snowballed into full-blown state corruption. What fixed it: ✅ Introduced remote backend with state locking (S3 + DynamoDB or GCS + locking table). ✅ Serialized CI/CD Terraform stages — no parallel applies on the same workspace. ✅ Added terraform plan validation before every apply. ✅ Regularly ran terraform refresh to sync with the actual cloud state. Result — ZERO “resource already exists” errors since. Terraform isn’t just code — it’s shared state with consequences. Treat it like a database, not a script. 🔁 Repost if you’ve ever chased a “random” Terraform error that turned out to be state-related. #Terraform #DevOps #CICD #InfrastructureAsCode #CloudAutomation #CloudComputing #SRE #GitOps #EngineeringLeadership #CloudNative #IaC

5