Najma Javed

When Android boots, it starts one primary process called the Zygote. This process pre-loads all the massive system libraries and resources (like the Android Framework and Dalvik/ART classes) that every app needs. fork() – The "Cheap" Copy When you launch an app like LinkedIn, the System Server doesn't create a new process from nothing. It tells the Zygote to fork() itself. The Magic of Copy-on-Write (COW). This is where the engineering gets brilliant. Under the hood, fork() doesn't actually copy the memory. It only copies the Page Tables (the maps that point to the memory). Shared State: Both the Zygote (Parent) and your New App (Child) point to the exact same physical memory pages. Read-Only: The kernel marks these pages as read-only. As long as the app is just "reading" system resources, no memory is duplicated. What happens when the app writes? The moment your app tries to modify a piece of memory (like setting its own unique variables): * The CPU raises a page fault. * The Kernel intervenes, allocates a new physical page for that specific data. * It copies only that tiny piece of data for the child. This is Copy-on-Write. #AndroidDev #OperatingSystems #SystemDesign #MobileEngineering #CodingTips #aaos #aosp #androidautomotive #androidframework #androidmiddleware #sairamkrishna #mammahe

28

1 comentário

It’s a quick deployment guide for Kubeview, a powerful visualization tool for Kubernetes clusters. Whether you're managing dev/test environments or optimizing multi-cluster views, Kubeview can offer real-time insights into your workloads. If you're working with Kubernetes and want a clean, visual way to track resources, this guide is for you. Feedback and shares welcome! #Kubernetes #DevOps #CloudNative #KubeView #Monitoring #SRE #PlatformEngineering

5

Spec-Driven Development: Inspired by GitHub Spec Kit   One of the biggest challenges in software development is unclear requirements and scope creep. Recently, I came across GitHub Spec Kit, and it reinforced an approach I truly believe in: Spec-Driven Development (SDD).   The idea is simple but powerful—start with clear specifications before writing any code. This structured workflow ensures alignment, reduces rework, and makes delivery predictable.   Here’s the flow that Spec Kit promotes: ✔ Specify – Define and validate actionable specs ✔ Plan – Outline architecture, tech stack, and dependencies ✔ Tasks – Break down into manageable steps ✔ Implement – Build and validate against the spec   This methodology isn’t just about process—it’s about mindset. If you’re curious, check out GitHub Spec Kit and see how it can fit into your workflow. Ref: https://lnkd.in/gAXt2pch #SoftwareDevelopment #SpecDrivenDevelopment #GitHubSpecKit #EngineeringBestPractices #ProjectManagement #TechLeadership #Agile #DevOps #ArtificialIntelligence #AgenticAI #AIAgent #AIExplained #TechTrends #Innovation #FutureOfAI #AIForBusiness

8

When Microsoft asks its own engineers to test a competitor against GitHub Copilot, the market should take note 👀 ​Thousands of engineers across Windows and Office are now trialling Claude Code. The signal is loud and clear: Productivity > Brand Loyalty. ​If a tool has superior agentic capabilities—like multi-step planning—it doesn't matter who owns it. The best talent wants the tool that helps them ship faster. ​The landscape is shifting weekly. We aren't seeing one dominant player anymore; we're seeing smart engineers building diverse toolkits. What is actually changing your workflow right now? Is it Copilot, Claude Code, Cursor, or something else? ​#Engineering #TechTrends #Recruitment #Productivity #AI #DevTools

This detailed infographic provides a simplified, holistic framework for modern web development, mapping the essential tech stack from foundational languages to core patterns and frameworks. It illustrates how the structural content of HTML and the styling of CSS are powered by the browser's core logic and scripting language, JavaScript. TypeScript is shown as a critical addition, providing a typed superset for scalable, maintainable code. The central Model-View-Controller (MVC) pattern acts as the architectural blueprint, separating data (Model), UI (View), and logic (Controller). The diagram then branches into popular choices: React, a library for building component-based UIs, and Angular, a full-featured application framework. A defining comparison is made on how these tools interact with the all-important Document Object Model (DOM). React utilizes a Virtual DOM abstraction for efficient updates, while Angular performs direct DOM manipulation, both using JavaScript to render the final web interface.

8

Any and all dependencies left in our code will come back with a vengeance when trying to get the code under test. All our modules are tightly coupled to a relatively large data structure that gets generated in the GUI or programmatically through a user-facing API for defining the actors, components, events and conditions in a simulation experiment.   These dependencies are weakening our codebase: they make changing the code difficult, it is not easy to read, and the underlying architecture fades away with so much noise.   The development team does not do Domain-Driven Design #DDD, yet the strategic design tools can be used to explain the issue and to see the value of the solution. We have 2 Bounded Contexts [Evans] tightly coupled together through a large data structure. The 2 Bounded Contexts can be described as: - Defining a simulation scenario (actors, sensors, environment, actions). This is where the large data structure is produced (upstream). - Generating a MATLAB Simulink model to run the simulation. This is the code we have been getting under test, where the large data structure is consumed (downstream).   The solution is simple. From the words of Martin Fowler: "I can easily rename or move a function while keeping the old function intact as a forwarding function (so my old code calls the old function, which calls the new function). ... Data is more awkward because I can't do that." [Refactoring] We are applying the "Encapsulate Variable" [Refactoring] to remove the coupling. We are moving the data structure behind an interface and step-by-step adding methods needed by the model creation process. Old ways of accessing the data structure remain, while progressively the new interface gets put in place on all the affected modules.   It has been surprisingly easy to adapt the existing tests to the new interface. The implementation details of dealing directly with a data structure get moved behind the new layer and our codebase looks better and better.   If you look at the developers working in #pairprogramming sessions, they seem to be just Encapsulating Variables again and again. The new interface acts as a seam [WEwLC] and shields the model creation code from an external ill-defined interface. What we are actually doing is refactoring toward deeper insight: improving the interface between 2 Bounded Contexts [DDD Blue Book].   This is work in progress. Once the whole structure is behind an interface, we could add Contract Tests Jan Van Ryswyck to have automated checks between the Bounded Contexts and let the teams work more autonomously and with a safety net.   #domaindrivendesign #ddd #tdd #refactoring #softwaredevelopment #softwarearchitecture #technology

12

Aren’t you bored of LLM posts? Here’s another one. ⬇️ https://lnkd.in/dxxf4Vg7 TLDR; 1️⃣ Create a well-defined schema for the output. Most frontier models have a dedicated JSON mode. This can be further fortified with Pydantic and/or Zod schemas to ensure that we don’t let malformed data through. 2️⃣ Use (at least) 2 LLM models to process the input asynchronously. You can even call the same model, but my intuition is that different models produce better results. This will dramatically decrease the rate of false positives (where the model thinks it extracted the correct data, but hasn't) and increase the overall success rate of the extraction. 3️⃣ Employ a retry logic. If the models don’t agree on the first iteration, we store their respective answers and feed them back with the original input. This will greatly enrich their context, already increasing the success rate by 20-30%. And if they fail for the second time… enlist the 'big guns' (i.e., top reasoning models). 4️⃣ LOG EVERYTHING! Every interaction with the models is recorded. I don’t think that at this point in time we are ready for fully autonomous agents that handle critical business data. LLMs will make mistakes even with these guardrails in place. We have to keep the responsible human in the loop and provide them with the necessary information to handle any “unfinished business”. #AI #DataExtraction #LLM #Automation #OCR #MachineLearning #GPT #DataScience #Innovation #Framework

5

1 comentário

The ONE mistake I made as a Software Engineer that made me rethink my whole career! 🙆🏽‍♀️ Sharing to encourage someone out there that no one is perfect or ✨successful ✨ without their fair share of knockouts. ► Once upon a time, I almost broke production code because of two letters in my code change (a unit addition) 😭. Had multiple UI + dimension changes on my PR, comprehensively tested everything (unit, espresso)… and just before I committed, I added this extra unit assuming I unintentionally missed it (this was the WRONGEST move). 🪄 Luckily, my senior teammate caught it during final integration testing because the app wasn’t launching after my changes 😭🤷🏽‍♀️🤸🏾‍♀️🙅🏽‍♀️. Before that, I couldn’t believe my change broke the app because I had thoroughly tested. Showed my test results + demos of everything working fine. BUT…. The logs don’t lie 🤣🤣. We went through the logs together and BOOM 💥 …. there it was, the little unit change I added last minute thinking nothing of it. _____ If I had retested after that change, I would have easily caught it 🤦🏾‍♀️. The lesson here is that even the littlest, simplest, and most infinitesimally small change can have MAJOR impact in a system. So, Pay attention and no matter what, test after every change. Even if you tested, test again and some more! More importantly, learn from your mistakes and GROW from them 🫶🏾. I talk more about failures in my latest YouTube video, go check it out and subscribe 👉🏾 https://lnkd.in/g8HRjyap

30

7 comentários

Highly recommended for anyone who has at least tiny bit of interface to Software Development, You can be a system engineer, safety manager, Test engineer, project manager etc. But it is very vital for anyone to understand how the foundation is SW development set, and how other activities in V cycle enable it. Save your spot before it is sold out!

7

Turning Vibe Coding Into a Live App: Token Length Checker Plus (Gemini) 🚀 Just shipped — a Gemini-powered web app that helps in: ✂️ Summarize long text 🧩 Simplify complex paragraphs 🔎 Extract Keywords & Hashtags 🧮 Estimate ≤ token length (great for LLM prompts/posts) 📝 Use an interactive quiz card to self-check understanding GitHub Repo: https://lnkd.in/gduSMZ7K 💡 Key learnings Spinning up an Ubuntu EC2 instance and running a Flask app on port 5000 Setting up a clean virtual environment and installing dependencies Configuring security groups so the app is reachable during testing Using system packages effectively (Python, pip, basic Linux tooling) 🌟 How it can be useful Students & teachers: Turn dense notes into quick summaries; quiz yourself to retain more Creators & marketers: Pull keywords/hashtags instantly and make content easier to read Engineers & PMs: Trim specs into concise briefs; keep drafts under token limits Anyone writing online: Get fast, clear versions before sharing/posting 🎯 Motivation & use I wanted a simple, everyday writing helper that’s fast, clean, and modular: Card-based UI with an interactive quiz, keeping raw JSON out of the way Add-on style architecture so features evolve without cluttering the core 🙏 Special thanks to Rohan Shravan for giving me the push to actually post on LinkedIn  #Flask #Python #Gemini #LLM #AWS #EC2 #LearningInPublic P.S : 🚧 Work in progress: polishing the UI and adding small UX tweaks next. ✨ Iterating fast—this is v1; more refinements coming soon. 🛠️ Still evolving: open to feedback before I lock v2.

29

AI will make your teams faster. It will not make your structure stronger. Most SDV programs measure velocity: release cadence, feature throughput, OTA frequency. Very few measure absorption capacity. But every engineering system runs on two rates: – the rate of change generation – the rate of change absorption If generation exceeds absorption, instability accumulates. Not because of talent. Not because of tooling. Because structure cannot absorb propagation. When change moves faster than structure, cost compounds. In my latest Substack article, I explore why absorption capacity — not velocity — determines the economics of software-centred vehicles. If your cost-of-change curve is steepening, the constraint may not be where you think it is. Full article below. #SoftwareCentredVehicle #SDV #AutomotiveLeadership #PlatformStrategy #OTA

8

1 comentário

Remember -- it's okay to ask questions as a software engineer. Effective product folks will work to enable success, whatever form that takes. 🧑‍💻 It's okay to: - ask questions about the support request - provide your realistic availability An effective product person will: - drive the open questions to zero to enable you and drive issues in specs and design documents to resolved and fleshed out - be willing to take partial work or accept an extended timeline or find someone else that can step in I had excellent engineering managers in my career that really helped me to internalize this. It's not "no, I can't do this", but instead a legitimate "I'd like to support you -- here's my questions and availability constraints". And don't say that and don't mean it; people can immediate smell it. Keep the vibes positive and constructive. At the end of the day we're all just "people trying to do stuff for business objectives". :)

13

1 comentário

What are Smart Pointers in C++ and Why Do We Need Them? Why not just use raw pointers? Smart pointers in C++ are objects that manage the lifetime of dynamically allocated memory. They automatically release memory when it is no longer needed — reducing the risk of memory leaks and dangling pointers. Types of smart pointers: std::unique_ptr → sole ownership std::shared_ptr → shared ownership (reference counted) std::weak_ptr → non-owning observer (used with shared_ptr) 🤔 Why use Smart Pointers when we already have raw pointers? 1. Automatic memory management: Raw pointers require manual delete, which can cause memory leaks if forgotten. Smart pointers call delete automatically. 2. Exception safety: If an exception occurs, smart pointers ensure proper cleanup. Raw pointers can easily leak memory in such cases. 3. Avoid common pointer bugs: Dangling pointers, double deletes, and memory leaks are reduced when using smart pointers. 4. Ownership semantics: Smart pointers make ownership clear (unique_ptr vs shared_ptr). Raw pointers provide no such clarity. Summary: ✅ Smart pointers = automatic, safer, and cleaner memory management. ❌ Raw pointers = powerful but error-prone if not carefully managed.

14

Let’s Dance! Reuse of cliches is a core of the System Architecture Discipline. Tango is completely different – it is a System-of-Systems built out of cliches, but executed with pure thought instead of blind routine. I’ve been practising tango for a while and now I’m ready to share my observations on the System Architecture behind this beautiful dance. According to the Systems Engineering Discipline, an Engineered System is built or modified for the human needs. Tango fits perfectly here. It is a social system with distinct and recognizable functions and elements – forming an emergent behaviour unreachable with the functions alone. Modern tango is an emergent System-of-Systems per se. It cleanly decomposes into the System of Music, System of Dance, System of Location, and a Social System. For instance, the System of Music breaks down into Tango, Valtz, and Milonga Sub-Systems. The System of Dance holds Global and Local Routing, Navigation, and Motion Sub-Systems. Finally, the Social System includes Etiquette, Relational, and Human Sub-Systems. What’s more – these systems are tightly interconnected and bring clear, distinct interfaces between the parts. This is what makes Tango particularly interesting for the System Architect. It is a naturally evolved system. Started as simple music, it bounced over the Atlantic Ocean twice – from Spain to Argentina and back – changing the world and conquering Europe and beyond. So what can a System Architect (and Engineering executive) learn from it? First of all, it operates with minimal friction when executed right. Guided primarily by the Music and Etiquette, it allows us to execute otherwise overly complex social interactions. It does not need the PPTs to identify global and local interfaces. It brings unprecedented flexibility to the execution of system behaviour. Because Music and Etiquette drive the architecture, the physical dancing itself becomes pure engineering of motion. The system seamlessly works in unpaved yards, saloons, outdoor terraces, or the great fests. It has a universal and low barrier DSL to communicate needs, desires, and actions. In the end, it brings true Leadership into the execution. Tango puts Leadership to its extreme. The Leader and the Follower are the persons. They are the Systems from the Engineering point of view. The dancing pair is a natural System-of-Systems, talking to each other, receiving, accepting, executing or sometimes objecting and declining impulses and responses. It is a natural and beautifully executed control with an active feedback loop. The Leader there not only assigns, but carefully listens to the follower and aligns own motion with the Follower. Tango embodies 'perfection' and 'beauty' in Engineering. It proves we must learn from established systems before building our own. Have live systems ever changed your way of thinking?

18

Functional Safety in Automotive: Because Every Line of Code Protects a Life I recently had the privilege of delivering an in-depth ISO 26262 Functional Safety training for the engineering teams at Mercedes-Benz — and it was an incredible experience to interact with some of the finest minds driving the future of mobility. As vehicles evolve into intelligent electronic systems, safety has become software’s biggest responsibility. With dozens of ECUs, millions of lines of code, and data-driven control — one unexpected fault can lead to a critical hazard. That’s where Functional Safety steps in. What ISO 26262 brings to the engineering table: ✔ Predictive risk analysis (HARA & ASIL) ✔ Systematic fault handling & safety mechanisms ✔ Hardware–software co-engineering for safety ✔ Traceability across the lifecycle ✔ Rigorous validation & MC/DC testing ✔ Confidence — even when things go wrong Embedded engineers today must build not just working software… but fault-tolerant, fail-safe, life-protecting systems. Why this matters more than ever: Autonomous drive Electric powertrains Smart braking & steering Safety-critical software is shaping the future of automotive innovation. I’m grateful for the opportunity to contribute to a safer mobility landscape — and excited to support more teams and organizations in their Functional Safety journey. Let’s build cars that think smart and react safely. #FunctionalSafty #ISO26262 #AutomotiveSafety #AutomotiveEngineering #SafetyCritical #EV #AutonomousVehicles #SoftwareSafety

86

1 comentário

🚗 Fresh from IAA Mobility 2025 – Munich! 🚗 Living in Munich, I sometimes forget I’m in the heart of automotive innovation—until weeks like this one. #IAAMobility 2025 reminded me just how fast the future is arriving! 💡 What stood out: 1️⃣ Mobility as experience – test drives, autonomous shuttles, and micromobility let people touch the future. 2️⃣ Global collaboration – 55% of exhibitors came from outside Germany. The race is global. 3️⃣ EVs + software = standard – electrification, software-defined vehicles, and sustainability are no longer optional. 4️⃣ Policy matters – infrastructure, regulations, and tariffs are shaping the pace of adoption. 5️⃣ Beyond cars – cycling, shared mobility, and public transport are all part of the mobility ecosystem. ✨ Personal highlights: • 🔋BMW iX3 “Neue Klasse” – Experiencing the #BMW iX3 NeueKlasse platform world premiere ,next-gen EV with circular design principles on Munich streets—and proud to have contributed to testing it as a #SystemEngineer for ADAS #IPNext! Seeing our work come alive in real-world demos was incredible. BMW Group • #MercedesBenz #CLA with #L2++ #ADAS • 🚙 #Cupra Tindaya Concept – proving design and brand identity matter as much as range and charging. • 🇨🇳 Chinese challengers everywhere – Leapmotor C10/B10, XPeng G6/G9 with 800V fast-charging, AVATR luxury EVs. Value, tech, and speed are reshaping the competitive map. • 🌱 Sustainability mainstreamed – circular materials, greener event concepts, and design thinking for the planet. • Experiencing #BYD EVs and learning how Chinese EVs are shaping the global market • Trying the #S_Bahn driving simulator—a fascinating glimpse into rail automation and urban mobility solutions. 🛴 Beyond EVs – Munich may be my living place but This city itself became a mobility test track with cycling, micromobility, and smart city solutions ,it feels like the epicenter of the mobility revolution! 🚀 🔑 Takeaway: Mobility 2025 isn’t just electric—it’s experiential, digital, and connected. The companies winning are those mastering speed, collaboration, and ecosystem thinking. #IAAMobility #Munich #FutureOfMobility #Innovation #SmartMobility #EV #ADAS #ElectricVehicles #Sustainability #MobilityEcosystem #SystemEngineering #IPNext #bmw #Mercedes #automotive #Germany #BYD IAA MOBILITY IAA

16

1 comentário