Addy Osmani is an Irish Software Engineer and leader currently working as a director at Google Cloud AI, focused on helping developers and businesses succeed with Gemini, Vertex AI, and the Agent Development Kit (ADK). After nearly 14 years leading developer experiences in Chrome, he now bridges Google DeepMind, engineering, product, and developer relations teams. A developer for 25+ years at Google, he is passionate about AI-assisted engineering and developer tools. Follow for motivation, productivity and engineering.

Addy Osmani

The real engineering skill isn't getting AI to 100% - it's recognizing when you've hit the point of diminishing returns at 90%. That's your cue to take the handoff.

You're not "giving up" by reading the code and finishing it yourself. 

You're doing what senior engineers do best: knowing when to switch tools, when to own the architecture and quality, and when speed matters less than understanding what you're shipping.

Let the AI sprint. You close with precision. It's okay if that last mile takes a little longer.

#ai #programming #softwareengineering

Director, Google Cloud AI. Best-selling Author. Speaker. AI, DX, UX. I want to see you win.

Chrome DevTools now supports *individual* network request throttling!

Folks have been asking for this for years! DevTools now allows developers to simulate slow network conditions for specific requests rather than the entire page. This helps in testing how a web application performs and handles issues when specific resources (like images, scripts, or API calls) are slow to load

Open the panel from ⋮ > More Tools > Request Conditions. Toggle the Enable blocking and throttling checkbox to enable request blocking. You can click the + button to add a new text pattern. 

#webdevelopers #programming #softwareengineering

Announcing big changes to Gemini in Chrome! Agentic browsing arrives.

Today marks a massive shift in how we experience the web. Starting today, Gemini in Google Chrome is evolving from an assistant into an agent that can navigate, reason, and execute complex tasks on your behalf.  

This includes a new design, Nano Banana support, deeper integration with Google products and agentic features like auto-browsing.

Here is what's changing:

👉 1. The Agentic "auto browse" 🤖

Gemini can now handle multi-step workflows across the web. Whether it's finding a walking tour on Expedia that fits your weather forecast or scrubbing your Redfin favorites for pet-friendly apartments and inviting your roommate to collaborate - Gemini does the clicking so you don't have to.

👉 2. "Nano Banana" image integration 🍌

Real-time image transformation is now native to Chrome. Powered by Nano Banana, you can now re-image furniture in a room or edit photos in-line without the tedious download-edit-upload cycle.

👉 3. Uninterrupted workflow with the new side panel 💻

Multitasking has a new home. The redesigned side panel stays with you as you browse, giving Gemini native access to your tabs. It can synthesize data from multiple sources, draft emails in Gmail based on a syllabus you’re reading, and even fill out complex registration forms from a PDF.

👉 4. Agentic commerce (Etsy, Expedia, & more) 🛍️

Planning a party? Show Gemini a photo and it will hunt down the supplies on Etsy, stay under your budget, and populate your cart. It’s moving from "searching" to "getting things done."

Rolling out first for Google AI Pro and Ultra subscribers in the US. 

#ai #programming #softwareengineering

Announcing Web Quality Skills: let AI agents write higher-quality code! 🚀

I'm excited to share Web Quality Skills - a comprehensive collection of Agent Skills designed to teach your AI coding assistants (Claude Code, Cursor, OpenCode, Antigravity) not just what to build, but how to build it right.

Repo: https://lnkd.in/gfGDWXTW
Skills: https://lnkd.in/g-Rrfn38

While most prompts focus on functionality, these skills encode the collective wisdom of web performance engineering into your AI.

What's inside?

✅ Guidance for many Lighthouse Audits (Performance, A11y, SEO)
✅ Core Web Vitals optimization patterns (LCP, INP, CLS)
✅ WCAG 2.1 accessibility standards
✅ Stack-agnostic (Works with React, Next.js, Vue, Astro, or plain HTML)

Instead of fixing performance after you deploy, you can now tell your agent:

"Audit this page for web quality issues" or "Optimize performance and fix Core Web Vitals"

It's like pair-programming with a performance expert. This is an unofficial project and something I personally use on my apps, but wanted to share it with y'all in case you find it useful. You can also pair this with the Chrome DevTools MCP so let your AI agent "see" what you do in the browser.

Get started with this one-liner:

npx skills add addyosmani/web-quality-skills

I hope you find it helpful! 

Demo in the video.

#ai #programming #softwareengineering

"Can you read 900 words per minute?"

RSVP is a way of displaying information, like words, one at a time, very quickly, in the exact same spot on a screen.

Instead of your eyes moving across a page to read a paragraph, your eyes stay fixed in one place while the content "flashes" by. It is essentially speed reading, where the content moves so your gaze doesn't have to.

#programming #softwareengineering

"The next two years of software engineering" five questions to ask yourself and your team with recommendations to consider

My latest free write-up: https://lnkd.in/g4EU-f5Z ✍ 

Software engineering is at a strange inflection point.

AI coding has evolved from autocomplete to autonomous agents. The hiring boom gave way to efficiency mandates. And a new generation of developers is entering the workforce having used AI assistance from day one.

I wrote about five questions that may shape our industry through 2026:

1. The junior developer question: Will entry-level hiring collapse as AI automates junior tasks or rebound as software spreads into every industry?

2. The skills question: Will core programming skills atrophy as AI writes most code or become more critical as humans focus on oversight and the hardest 20%?

3. The role question: Will developers shrink into code auditors or expand into orchestrators designing AI-driven systems?

4. The specialist vs. generalist question: Will narrow specialists find their niche automated or will T-shaped engineers who blend depth with breadth thrive?

5. The education question: Will CS degrees remain the gold standard or will faster paths (bootcamps, portfolios, employer training) overtake them?

For each question, I explore two contrasting scenarios and offer concrete recommendations for both junior and senior developers.

The consistent thread: change is the only constant. I hope some of the ideas here help you stay in the loop regardless of which future arrives.

#ai #programming #softwareengineering

Vibe Kanban: orchestrate multiple AI coding agents in parallel. Free and 100% open-source!

Switch between Claude Code, Codex Gemini CLI, and track task status from a single dashboard. Repo: Repo: https://lnkd.in/gA_UPD8a

Vibe Kanban is essentially a project management layer for your coding agents. Instead of just staring at a terminal while Claude Code or Gemini CLI does its thing, this gives you a UI to:

1. Orchestrate multiple agents: You can run them in parallel or sequence.
Visualise the work: It puts tasks on a board so you can actually track status (hence the Kanban name).

2. Centralise config: Manage your MCP configurations in one spot.

3. Remote access: It handles SSH, so you can open projects on remote servers easily.

I've been enjoying trying out a few orchestration tools this last year, including Conductor, Jules, Claude Squad and of course Claude Code Web, GitHub Copilot Agent etc. While the tools you use might change, I think now is a great time to build your skills and comfort levels using parallel agents. 

#ai #programming #softwareengineering

Build a full-stack Deep Research Agent with React and Google ADK!

Research is one of the most popular use-cases folks have for agents. 

Today I'd like to show you our Deep Search Google ADK (Agent Development Kit) sample - a sophisticated, full-stack application that solves two major problems in AI development: lazy execution and misalignment.

It plans before it acts: instead of rushing to search, it pauses to strategize. it presents a research plan to the user first, allowing for a "human-in-the-loop" moment to refine the direction before resources are spent.

It self-corrects: it implements a "critic" loop. if the research findings are weak or gaps are found, the agent autonomously generates new questions and searches again until the quality bar is met.

This is also not just a notebook: this is a complete app including a React + Vite frontend, a Python + FastAPI backend, and deployment scripts for Google Cloud, making it a true template for production apps. Even if your Python is dusty, I promise you'll find it relatively straight-forward to navigate.

Check out the GitHub repo: https://lnkd.in/gbAe5Zck

...as well as the many, many agent samples we've got in the same place :)

#ai #programming #softwareengineering

"How to write a good spec for AI agents"

My latest free write-up: https://lnkd.in/g--r4kEu  ✍ 

Using an AI agent for coding or other use-cases? I keep hearing "just write a better spec" but the real problem isn't effort. You can write something that rivals an RFC and still watch the model fall apart once the context gets too large.

What's worked for me is aiming for specs that are clear, scoped, and designed for how agents actually process information, not how humans wish they would.

A few principles that consistently pay off:

Start with a high-level vision, then let the agent draft the details. Treat the first pass like a product brief. Review, correct, and refine before any code is written.

Structure the spec like a real PRD. Clear sections, explicit commands, testing expectations, style examples, and hard boundaries beat loose prose every time.

Break work into modular tasks instead of one giant prompt. Small, focused context wins over dumping everything into a single message.

Build in guardrails. Define "always do,” "ask first,” and "never do.” Encourage self-checks and verification against the spec before you review anything.

Plan first, execute second. Use read-only or plan modes where possible, then iterate. Specs work best as living documents that evolve with the project.

The takeaway: good agent specs aren't about writing more; they're about writing just enough structure and nuance to keep the model focused without overwhelming it.

If you've been frustrated by agents drifting, hallucinating, or missing requirements, the fix is usually upstream. The spec is the first thing you and the AI build together.

#ai #programming #softwareengineer

"My AI coding workflow in 2026"

Write-up: https://lnkd.in/gvcs_5sc ✍ 

I get asked about my workflow for coding with LLMs pretty often so I thought I would share some notes on what works well for me. 

Here is my 5-step workflow for this year:

1. Specs before code (The "15-Minute waterfall") 

Don't dive straight into generating code - plan. I start by brainstorming with the AI to create a comprehensive spec.md and a step-by-step project plan. We define requirements, UX, and edge cases before a single line of code is generated. Really improves quality. Cursor, Cline, Claude Code and many tools now encourage more planning before acting to support this workflow.

2. Context is king - without it, LLMs can make incorrect assumptions.

LLMs are only as good as the information you feed them. I use "context packing" - dumping relevant docs, existing patterns, and constraints into the session or project directory (using tools like gitingest or Context7). Don't make the AI guess or waste tokens trying to figure out what's in your head.

3. Manage task granularity - just like with traditional engineering, do work in well defined tasks or chunks, not all at once.

Break work into iterative chunks. "Implement one function/component" works infinitely better than "Build this app." I use a structured prompt plan, executing tasks one by one to keep the model on the rails. 

4. Review and test what's important - LLMs are often an "over-confident pair programmer" 

It's healthy to treat AI output like code that came from an over-confident pair developer. It might look right, but is it? Review and test everything important. I weave testing into the loop - generating tests alongside the code and refusing to merge until they pass. If testing feels too much for your project (maybe it's a personal project), at least review the diffs. 

5. Git is your save point - commit early and often. 

When an AI hallucinates or breaks the build (and it will), you need granular commit history to roll back to a safe state immediately. I've had so many experiences where I feel great 3 prompts in then the next 2 completely go off the rails. Tools increasingly support the idea of checkpoints and rollbacks, but for most workflows, just using git + committing can get you far.

In terms of models and tools: I have a core set of models (incl Gemini 3) and tools that I rely on pretty regularly, but I make time each week to try out what's new (patterns, open-source projects etc). 

So if the Ralph Wiggum loop, Gastown/Beads or other ideas trend they're on my radar and I can choose whether they make sense for my workflow.  I'm very big on orchestration right now and use CLIs for more complex local work. It's very much just continuous learning + trying things.

AI tools are incredible force multipliers, but they don't replace the need for engineering fundamentals. In fact, they reward them. The human engineer remains the director of the show.

#ai #programming #softwareengineering

Announcing "Product Engineering with AI" - a practical, 200-page book bundle on building AI-powered products from foundations → tools → real-world implementation!

Get it (and sample chapters) at: https://prodengineer.org  🚀 

If you're a software engineer or product manager trying to ship with agents, this is the playbook we wish we had. Shipping AI features today is less "write every line" and more "steer tools + editors + agents, then verify quality."

Hassan Djirdeh and I wrote Product Engineering with AI to make that shift concrete: AI dev platforms, AI code editors, and agentic workflows - plus prompt patterns, debugging, and UX practices that hold up beyond prototypes.

The bundle also includes bonuses:

• Product Engineering with Claude Code: hands-on PM workflows (parallel agents, commands, MCP integrations). Lots that also apply to Gemini CLI.
• AI in Practice: a 30-page case study collection of real production AI examples
• Product Engineering Insights in 2026: insights from experts across the field on how roles, tools, and workflows are changing
• Google Agent Development Kit (ADK) samples - 5 new agent code examples for product engineers including a PRD Studio, Customer & Meeting analysis and more.
• A 45-min workshop: how we built an AI-powered product from scratch
• A 30-min talk: designing human-centered AI experiences people trust

We hope you find it useful!

#ai #programming #softwareengineering

"The AI coding trends for 2026" (so far!)

My latest free write-up: https://lnkd.in/gjFFq_8N ✍ 

The AI coding landscape in 2026 has shifted fast. I just published a primer on the patterns and platforms defining how we build software this year. It's going to keep evolving and I plan to keep my write-up updated.

Here are the top trends shaping 2026:

🔄 The Ralph Wiggum pattern: We are moving from single-shot prompts to autonomous loops. Agents now run tests, hit errors, and fix their own code until the "completion tag" lights up. 

📦 Agent Skills: Expertise is becoming portable. Just like npm install, you can now install "skills" (like Vercel's performance rules or accessibility guidelines) directly into your agent's context.

🏗️ Orchestration > Assistance: Tools like Conductor, Vibe Kanban, and Gas Town are turning developers into fleet commanders. We aren't just pairing with one AI. We are managing parallel agents working in isolated git worktrees. I see this becoming a more default way we work.

🧩 Sub-Agents: Monolithic assistants are out. Specialized sub-agents (one for security, one for docs, one for tests) are in, reducing context pollution and hallucinations.

🦞 We are also seeing a massive rise in local-first tools like Clawdbot (with a lot of hype around it right now) and how we approach personal intelligence. 

Even at Google, we know that the industry feels like it's moving fast and I wanted to try making it just a little easier to stay updated as we all share and figure this stuff out :) Also included my fav. community posts as a bonus.

#ai #programming #softwareengineering #lifeatgoogle

Full Stack developer | content creator| open for collabs| freelancer

🚀 Relentlessly Growing 1% Every Day

I’m a Senior Backend Developer currently working at 42Gears, where I focus on
building scalable, reliable, and high-performance backend systems used at
enterprise scale.

Previously, I’ve worked in a startup environment at AstrotimeApp, which
strengthened my ability to move fast, take ownership, and deliver
production-ready solutions end-to-end.

My journey in tech is driven by continuous learning and real-world problem
solving. I believe in improving a little every day—whether it’s mastering a
new framework, refining system design, or optimizing performance in
distributed systems.

💻 Core Technical Expertise

Backend & Systems  
• Java, Spring Boot  
• Node.js, Express.js  
• .NET  
• Golang, Python  

Databases & Messaging  
• MongoDB, Microsoft SQL Server  
• Redis  
• Apache Kafka  
• MQTT  

Frontend (Working Knowledge)  
• AngularJS, Angular  
• TypeScript  

Cloud & DevOps  
• AWS  
• Docker  
• Microservices Architecture  
• Nginx  

🌱 What I’m Focused On Right Now  
• Designing and scaling microservices  
• Improving system reliability, performance, and observability  
• Deepening expertise in cloud-native architectures  
• Writing clean, maintainable, production-grade backend code  

🤝 Open to sharing learnings, career guidance, and real-world backend
experience with developers and students.

👉 Book a 1:1 career guidance session here:  
https://topmate.io/kartik_kaushik11

Kartik Kaushik

Hi LinkedIn family,

I’m writing this post with a lot of hope.

My sister has recently moved to the UK for her higher studies.
Like many international students, she is trying to stand on her own feet
and manage her living expenses while pursuing her dreams.

As her brother sitting miles away this is one of those moments where
you feel helpless  wanting to do more but limited by distance.
So today I’m turning to this community.

She is urgently looking for job opportunities in areas like
sales, marketing, retail, or any suitable entry-level roles.
She is hardworking, sincere, and ready to learn.

If anyone here can help with leads, referrals, guidance,
or even a small direction it would mean more to us than I can express.

I can share her LinkedIn profile and resume with anyone willing to help.

Thank you for taking a moment to read this.
Grateful for the power of this community.

#UKJobs #StudentJobsUK #HiringUK #Referrals #InternationalStudents
#SalesJobs #MarketingJobs #LinkedInCommunity

Senior Software Engineer | content creator| open for collabs| freelancer | 1:1 Career Guidance

Hi Linkedin Community 

A Great List of Tech Articles to Read This Weekend:

1)How Airbnb's improved latency by 150x on Transaction History page?
 👇 by Kamran Ali
https://lnkd.in/g3PnySvU

2)Coding Challenge #103 - Software Teleprompter by John Crickett:https://lnkd.in/g4Y5aNkh

3)Polling or Webhooks : By Saurabh Dashora
https://lnkd.in/gjTi-yXC

4)10 Algorithms Every Engineer Should Still Know in 2026: 👇:. By Alexandre Zajac
https://lnkd.in/gcBz7XGE

6)If you want to grow as a software engineer in 2026, by Nikki Siapno
https://lnkd.in/gpK35cFE

7)Tech Lead is one of the most desirable roles in tech.
? Gregor Ojstersek
https://lnkd.in/gQ2Gua8e

8) Understanding algorithms changes how you code. By Mayank A.:
https://lnkd.in/ggT6BrFb

9)MCP - A Deep Dive✨ By Neo Kim and Eric Roby:https://lnkd.in/e7De898X

10)Resilience doesn’t break inside services. By Raul Junco:https://lnkd.in/gYnscYwg

11)CAP Theorem Explained in simple terms:
: by Ashish Pratap Singh:https://lnkd.in/gH7u6NMH

12)Scaling monorepos does not mean rebuilding everything. . By Slobodan Gajić
https://lnkd.in/gEKxpurm

13)𝗠𝗮𝘅𝗶𝗺𝘂𝗺 𝗦𝘂𝗯𝗮𝗿𝗿𝗮𝘆 𝗦𝘂𝗺 - 𝗕𝗿𝘂𝘁𝗲 𝗙𝗼𝗿𝗰𝗲 𝘃𝘀 𝗞𝗮𝗱𝗮𝗻𝗲'𝘀 𝗔𝗹𝗴𝗼𝗿𝗶𝘁𝗵𝗺
: By NeetCode: https://lnkd.in/gdfKKFCE

14)If you think knowing about data structures and algorithms is not important in software development By Fernando Franco:https://lnkd.in/g-6euBmE

15) OAP vs REST vs GraphQL vs RPC. By Alex Xu

https://lnkd.in/gr9rNPeK

16)Stop obsessing over which LLM is better.
 by Nina Fernanda Durán
https://lnkd.in/guuCAEgU

17) If you’ve experimented with the new Angular Signal Forms API, ​🧪 🚀 Roberto Heckers
https://lnkd.in/gpZBCyXX

18)Imagine you have 1 million drivers in a database.   By sukhad anand
https://lnkd.in/guUs3zgt

19)Advent of Agents: 25 free days of building AI agents for production!
 By Addy Osmani: https://lnkd.in/gWX62yFM

20)Stop blaming others when you find a bug. By Daniel Moka: 
https://lnkd.in/g76cBHTa

21)🎯Someone hit retry when their network hiccupped. 
 Abhishek Kumar:https://lnkd.in/gQXDBaXZ

22)How to Push Back on Bad Ideas (Without Killing Your Career):.By Taha Hussain:https://lnkd.in/gDmrGa6H

23)If you want to improve your code reviews,. Petar Ivanov:https://lnkd.in/gdi59XhH

24) 40 Kubernetes concepts
 By Mahesh Mallikarjunaiah ↗️
https://lnkd.in/gdPRBpE4

25)How to Build Your First ChatGPT App By Jordan Cutler:https://lnkd.in/gfZ_2c5R

26)8 Tips To Write Clean Code by Milan Jovanović https://lnkd.in/g2yYeNbK

27)I’ve supported hundreds of engineers and leaders with various ambitious goals by Irina Stanescu
https://lnkd.in/gQFYcQcZ

28)𝗧𝗵𝗲 𝗚𝗼 𝗽𝗮𝘁𝘁𝗲𝗿𝗻 𝘁𝗵𝗮𝘁 𝗿𝗲𝗽𝗹𝗮𝗰𝗲𝘀 𝗿𝗲𝘁𝗿𝗶𝗲𝘀, 𝘁𝗶𝗺𝗲𝗼𝘂𝘁𝘀 𝗮𝗻𝗱 𝗰𝗶𝗿𝗰𝘂𝗶𝘁 𝗯𝗿𝗲𝗮𝗸𝗲𝗿𝘀, by Rihab SAKHRI:https://lnkd.in/gvBbYXHW

29)Truncate is a DDL By Hussein Nasser :https://lnkd.in/gcb9KtG4


Happy Weekend Everyone

Senior Software Engineer | content creator| open for collabs| freelancer

Hi Linkedin Community 

A Great List of Tech Articles to Read This Weekend:

1)How Uber built replication system to handle ~300,000,000 GB 
 👇 by Kamran Ali
https://lnkd.in/gMXJCnWu

2)Coding Challenge #105 - Top Programming Stories  by John Crickett:https://lnkd.in/g6eXKaPi

3)Sticky Sessions or Not? : By Saurabh Dashora
https://lnkd.in/gzmUk5zT

4)10 dead simple tips to start : 👇:. By Alexandre Zajac
https://lnkd.in/gYSAiyza

6)Vector Search vs Lexical Prefiltered , by Nikki Siapno
https://lnkd.in/gTGGhEaU

7)Coinbase scaled their hiring to 150 engineers per month over the pandemic.
? Gregor Ojstersek
https://lnkd.in/gFw4KMFu

8) Building AI agents is easy, getting them to survive production  By Mayank A.:
https://lnkd.in/gpSnXaAG

9)System Design Interview: Design Airbnb✨ By Neo Kim :https://lnkd.in/gqj3ufpF

10)The biggest lie in software architecture: By Raul Junco:https://lnkd.in/g9p8Sj6T

11)CAP Theorem Explained in simple terms:
: by Ashish Pratap Singh:https://lnkd.in/gH7u6NMH

12)Scaling monorepos does not mean rebuilding everything. . By Slobodan Gajić
https://lnkd.in/gEKxpurm

13)𝗦𝘂𝗯𝗮𝗿𝗿𝗮𝘆 𝘃𝘀 𝗦𝘂𝗯𝘀𝗲𝗾𝘂𝗲𝗻𝗰𝗲 𝘃𝘀 𝗦𝘂𝗯𝘀𝘁𝗿𝗶𝗻𝗴 𝘃𝘀 𝗦𝘂𝗯𝘀𝗲𝘁
: By NeetCode: https://lnkd.in/gJDPDBQQ

14)16 important DSA Problem Patterns you should look at while preparing for coding By Fernando Franco:https://lnkd.in/gkckTJww

15) Modern Storage Systems. By Alex Xu

https://lnkd.in/gbemmdDy

16)If you cannot build an AI agent without a library, you do not understand how it works.
 by Nina Fernanda Durán
https://lnkd.in/gGFcmvRr

17) Angular v21.1 is released and one of my favorite new features is support for the spread operator ​🧪 🚀 Roberto Heckers
https://lnkd.in/gsBWiBdD

18)In a Monolith, debugging is easy. By sukhad anand
https://lnkd.in/gCPUyb-T

19)"Can you read 900 words per minute?"
 By Addy Osmani: https://lnkd.in/gGPbQHb5

20)Career tip for 2026: Go all in on AI.. By Daniel Moka: 
https://lnkd.in/gp_D5svz

21)🎯I bombed my Google system design interview.
 Abhishek Kumar:https://lnkd.in/gHZpvABy

22)Trying to be cool is the fastest way to stay invisible.
:.By Taha Hussain:https://lnkd.in/g5TAA-J4

23)If a function accepts multiple, Petar Ivanov:https://lnkd.in/g6M5XWax

24) Here are 9 must read books if you have made a resolution to get serious about 
 By Mahesh Mallikarjunaiah ↗️
https://lnkd.in/gkgQXsAF

25)How to Build Your First ChatGPT App By Jordan Cutler:https://lnkd.in/gfZ_2c5R

26)Data Annotations are great for simple things. by Milan Jovanović https://lnkd.in/gYF4A9cH

27)CEO: "Don't be filtered. I want people to tell me what they really think." by Irina Stanescu
https://lnkd.in/gfEmc7y7

28)𝗦𝗹𝗶𝗰𝗲 𝗯𝘂𝗴𝘀 𝗮𝗿𝗲 𝘁𝗵𝗲 𝗺𝗼𝘀𝘁 𝗱𝗮𝗻𝗴𝗲𝗿𝗼𝘂𝘀 𝗯𝘂𝗴𝘀 𝗶𝗻 𝗚𝗼 by Rihab SAKHRI:https://lnkd.in/gJr9NM4P

29)TCP is a single stream By Hussein Nasser :https://lnkd.in/gDT9kRiN

Hi LinkedIn community,
 🙏 Jai Shree Krishan to everyone!

📩 Message Brokers 

 Let’s break it down section by section 👇

1)📨 What is a Message Queue?
At its core, a message queue sits between producers and consumers.
Producer sends messages
Message Queue stores them temporarily
Consumer processes them asynchronously

📌 Key benefit: decoupling
Producers don’t wait for consumers
Consumers can be slow or down
System remains stable
This is the foundation of resilient architectures.

2)🔔 Publish–Subscribe Pattern (Event-Driven Systems)
In the right side of the image, notice the New Order Topic.
Here’s what’s happening:
Order Service publishes an OrderCreated event
Multiple services subscribe:
Inventory Service
Shipping Service
Notification Service
📌 One event → multiple independent reactions
 📌 No direct service-to-service calls
This is how event-driven microservices scale without becoming tightly coupled.

3)🔄 Steps in a Messaging System (End-to-End Flow)
The middle section shows the full lifecycle of a message:

1️⃣ Production
Producer creates and sends a message.

2️⃣ Ingestion
Broker:
Stores the message
Replicates it for fault tolerance

3️⃣ Consumption
Consumer fetches the message from the broker.

4️⃣ Processing
Business logic is executed:
DB writes
API calls
Side effects

5️⃣ Acknowledgment
Consumer confirms successful processing.
📌 Until acknowledgment → message is not considered done.
⚖️ Competing Consumers Pattern (Horizontal Scaling)

The bottom section shows multiple consumers competing for messages.
All consumers read from the same queue
Each message is processed by only one consumer
Broker load-balances messages

📌 This enables:
Horizontal scaling
Parallel processing
Faster throughput
Add more consumers → system scales automatically.

🧠 Why this matters in real systems
This entire flow gives you:
Fault tolerance
Loose coupling
Scalability
Back-pressure handling
And this is why message brokers are used in:
Order processing
Payments
Notifications
Data pipelines
Event sourcing

🧩 Final takeaway
This diagram is not about queues or topics.
 It’s about designing systems that don’t break under load.
If you can explain:
Message queues
Pub-sub
Message lifecycle
Competing consumers

Image credits : Alex Xu (ByteByteGo)

#MessageBrokers #SystemDesign #DistributedSystems #EventDrivenArchitecture #BackendEngineering #Kafka #RabbitMQ #Scalability

Hi LinkedIn Community,
 Jai Shree Krishna to everyone 🙏

📌 How Pinterest Built an Async Compute Platform for Billions of Task Executions

Every time you save a Pin on Pinterest, it looks instant.
 Behind the scenes, however, billions of background jobs are being executed every single day — image processing, notifications, analytics, recommendations, and more.

This is the story of how Pinterest evolved from Pinlater → Pacer, and why it’s a gold-standard system design case study.

1)🔁 Why Async Processing Is Critical
Not all tasks need to run synchronously.
Pinterest immediately acknowledges user actions while pushing heavy work into background queues. This keeps the UI fast — but at massive scale, async systems become extremely hard to operate correctly.

2)⚠️ Where Pinlater Broke at Scale
Pinlater worked well initially, but growth exposed fundamental flaws:
🔸 Database lock contention
 Every queue was spread across all DB shards, forcing multiple servers to scan the same partitions concurrently.
🔸 No workload isolation
 CPU-heavy image processing jobs ran alongside lightweight notification jobs — one failure could take everything down.
🔸 Shared critical & non-critical paths
 Enqueue (user-facing) and dequeue (background) operations competed for the same resources.
🔸 Massive inefficiency
 Over 50% of DB queries returned empty results.
🔸 No FIFO guarantees
 Jobs were processed from multiple shards with no global ordering.
Scaling infrastructure only made things worse.

3)🚀 Pacer: A Ground-Up Redesign
Instead of incremental fixes, Pinterest rebuilt the system.

🧩 Key Architectural Innovations
🔹 Dedicated Dequeue Broker (Stateful Layer)
 Each queue partition is owned by exactly one broker, coordinated using Helix + ZooKeeper.
 👉 Lock contention disappeared completely.
🔹 In-Memory Job Buffers
 Brokers prefetch jobs and serve workers directly from memory.
 👉 Dequeue latency dropped to sub-millisecond (<1ms).
🔹 Adaptive Sharding
Small queues → few partitions
Large queues → many partitions
 👉 Empty DB scans dropped from 50% → nearly zero.

🔹 FIFO Ordering (Finally!)
 Single-partition queues now guarantee strict ordering when required.
🔹 Isolated Execution via Kubernetes
 Each queue runs in its own pod with custom CPU/memory/concurrency.
 Failures are contained. Performance tuning is finally possible.
🔹 Critical Path Isolation
Enqueue = user-facing & always reliable
Dequeue = backend processing
 Even if workers slow down, users never notice.

4)📈 Measurable Impact
✅ Lock contention eliminated
 ✅ Linear scalability by adding brokers
 ✅ Faster job execution
 ✅ Better hardware efficiency
 ✅ Zero cross-queue blast radius


#SystemDesign #DistributedSystems #Scalability #BackendEngineering #AsyncProcessing #Pinterest #TechArchitecture #EngineeringLeadership

Hi LinkedIn community 👋
 Jai Shree Krishna to everyone 🙏✨

Netflix is no longer just a streaming platform 🎬

 It now spans live events, mobile gaming 🎮, ads, and hyper-personalized experiences—and that evolution created one of the most fascinating real-time distributed systems challenges at internet scale.

1)🎯 The Core Problem
A single Netflix member might:
Watch Stranger Things on a smartphone 📱
Continue on a smart TV 📺
Launch the Stranger Things mobile game on a tablet 🎮
Different devices. Different services. Different times.
 Yet it’s all one member journey 🧩
Behind the scenes, Netflix runs hundreds of microservices, each owning its own data store.
 Streaming data, gaming data, authentication data—all lived in silos. Traditional data warehouses couldn’t connect these relationships in real time.

2)🧠 Why Netflix Chose a Graph
Because member journeys are relationships, not rows.
Graphs enabled Netflix to:
⚡ Traverse relationships without expensive joins
🔄 Adapt easily as new connections appeared
🔍 Detect patterns across devices, content, and behavior
This led to RDG — Real-Time Distributed Graph.

3)🏗️ How RDG Works (Simplified)
Kafka 📩 ingests millions of real-time events per second
Flink 🔄 processes, enriches, deduplicates, and converts events into nodes & edges
KVDAL (built on Cassandra) 🗄️ stores the graph using adjacency lists for ultra-fast traversal
Instead of relying on traditional graph databases (which struggled with scale and ingestion), Netflix built graph behavior on top of proven distributed systems.

4)🧩 Smart Design Decisions
One Kafka topic → one Flink job (simpler tuning & reliability)
Separate namespaces for each node and edge type
Adjacency lists for lightning-fast reads ⚡
Fine-grained TTLs to control graph growth
Simple ideas. Massive impact.

5)📈 The Scale (Production Numbers)
🟢 8+ billion nodes
🔵 150+ billion edges
⚡ ~2M reads/sec
✍️ ~6M writes/sec
⏱️ Single-digit millisecond latency
And the system scales linearly 🚀
🔑 The Big Takeaway
Netflix didn’t chase shiny technology.
 They optimized for reliability, operational simplicity, and real-time performance.
This is engineering pragmatism at internet scale 🧠🔥
If you’re building systems involving personalization, real-time context, or cross-platform user experiences, Netflix’s RDG is a blueprint worth studying.

#SystemDesign #DistributedSystems #NetflixEngineering #GraphArchitecture #BackendEngineering #DataEngineering #RealTimeSystems #Microservices 🚀

Hi LinkedIn community,
Jai Shree Krishna to everyone 🙏

🚀 We’re Hiring at 42Gears | Multiple Open Positions – Bangalore 🚀
At 42Gears, we build secure, scalable, and enterprise-grade Unified Endpoint Management (UEM) & Device Management solutions trusted by organizations worldwide.
We are expanding our teams across Engineering, Cloud, Security, QA, DevOps, UI, Sales, and Design.
If you’re passionate about technology, innovation, and solving real-world enterprise problems, we’d love to connect with you!

🌟 Why Join 42Gears?
✔ Work on global enterprise products used by customers worldwide
 ✔ Exposure to cutting-edge technologies in Cloud, Security, MDM & UEM
 ✔ Strong learning culture with mentorship from experienced leaders
 ✔ Opportunity to own impactful features end-to-end
 ✔ Career growth in a fast-scaling, product-based organization
 ✔ Collaborative, transparent, and high-performance work culture

🔹 Current Openings

A)👨‍💻 Engineering & Product Development

1)Senior Software Engineer (Windows / Device Platform) – 3–4 yrs
Skills: C#, .NET, WPF/XAML, WinForms, MVVM, Windows Internals

2)Senior Software Engineer (macOS / Apple Platform) – 3+ yrs
Skills: Swift, Objective-C, AppKit, LaunchAgents, XPC, macOS Security Frameworks

3)Software Developer (Java & Linux) – 0–5 yrs
Skills: Core Java, OOP, Design Patterns, JVM, Linux, Debugging

4)Software Engineer – Freshers (Java) – 0 yrs
Skills: Java, OOP, Design Patterns, Test Automation, Problem Solving

5)Senior UI Developer – 4–5+ yrs
Skills: Angular (v10+), TypeScript, RxJS, HTML5, CSS3, SCSS, REST APIs

B)☁️ Cloud & Infrastructure
a)Lead Cloud Engineer – 6–8 yrs
Skills: AWS, Azure, GCP, VPC/VNet, IAM, Security, Observability, DR, Cost Optimization

b)Associate Cloud Manager – 10+ yrs
Skills: Cloud Strategy, Governance, Hybrid Cloud, Team Leadership, Agile Delivery

c)🔧 DevOps & QA
DevOps Engineer / Jr. DevOps Engineer – 0–3 yrs
Skills: CI/CD (Jenkins/GitHub Actions), Docker, Kubernetes, AWS/GCP, Linux

d)Test Engineer (QA) – 1–3 yrs
Skills: SSO, IDP (Okta, Google, Microsoft Entra), OAuth, SAML, MDM Testing
🔐 Information Security

e)Senior Security Engineer I – 5+ yrs
Skills: VAPT, Thick-Client Security, Source Code Review
Senior Security Analyst – Malware & Application Security

f)🎨 Design
a)Senior Graphic Designer – 3–5 yrs
Skills: Adobe Photoshop, Illustrator, InDesign, Figma, Video Editing
💼 Sales
b)Channel Sales Manager – 8–10 yrs
Skills: Partner Management, SaaS Sales, Channel Enablement

c)Sales Executive – 0–2 yrs
Skills: Field Sales, Lead Generation, Partner Ecosystem

d)🎓 Internship
Intern – Professional Services – Freshers
Skills: SQL, SDLC, Customer Support, Product Deployment

📍 Location: Bangalore
🕘 Employment Type: Full-Time (Internship available)
📧 How to Apply:
👉 Please share your resume at: kartikgraphic6@gmail.com

📌 Mention  Role in the subject line (– Lead Cloud Engineer)


#Hiring #42Gears #BangaloreJobs #TechHiring #CloudJobs

taplio

The Battle

Addy Osmani

Kartik Kaushik

Followers

Posts

Engagements

Likes

Performance ScoreAddy Osmani's LinkedIn Performance

Performance ScoreKartik Kaushik's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Addy Osmani

Kartik Kaushik

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Addy Osmani's LinkedIn Performance

Performance ScoreKartik Kaushik's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreAddy Osmani's LinkedIn Performance