Ritesh Roushan

Two senior backend engineers reviewed the same system: Dev A: Added Kafka between services. Dev B: Merged three synchronous API calls into one. Same traffic. Same hardware. The API latency dropped by 80%. Whose design would you approve ?

@SahilPanhotra Okay okay, let's see others people explanation too. Because now a days asking such questions in interview from SDE 1 guy also

@devXritesh I'd use regional edge nodes with failover to keep latency under 200 ms : P

System Design Interview Question: You're building a Zoom-like video conferencing platform for enterprises. Requirements : Support 1 million concurrent participants globally. Meetings can have up to 500 participants. Users should automatically connect to the nearest region for low latency. If a regional data center fails, meetings must continue with minimal disruption. Support screen sharing, chat, recording, and real-time captions. Video quality should automatically adapt to changing network conditions. End-to-end latency should stay below 200 ms. How would you design the architecture? Explain your High-Level Design 👇

Tip: Break the problem into signaling, media, and infrastructure layers.

One thing I've observed over the last two years as a software engineer... In 2024, the narrative was: "AI will replace developers." "AI will replace designers." "AI will replace support." "AI will replace almost every tech role." Fast forward to 2026. The conversation has changed. The teams building the best products aren't replacing people with AI. They're helping people become more productive with AI. AI can generate code. It still can't fully understand your business context, make product trade-offs, debug complex production incidents, or take ownership of a system. The biggest shift I've noticed isn't in the technology. It's in how companies use it. Less "replace humans." More "augment humans." AI has become one of the most powerful tools we've ever had. And like every great tool, its value depends on the engineer using it. What have you observed over the last two years?

@GohilHardy That's fair thing, I like after AI based development

@devXritesh AI removed a lot of repetitive work, which is a good thing.

@GohilHardy Distribution is real issue for tech founder.

Signs a founder is probably stuck 👇🏻 - Can't explain the product clearly - Keeps building but never launches - Talks valuation before revenue - Avoids marketing completely - Underprices the product badly Most problems aren't technical. They are distribution problems.

@AlfinCodes 8/10 because I don't like messy stuff much

Rate this setup

Do share your thoughts here

@SourabhGurwani CDN one of the most important topics in distributed systems

System Design From Scratch — Day 8 CDN (Content Delivery Network) 1. Why? Users are worldwide. Fetching every image/video from one server = Slow. 2. Simple Explanation A CDN stores copies of your static files across the globe. Users download content from the nearest server, not the original one. 3. Real-World Example When you watch videos on Netflix or YouTube, most content comes from a nearby CDN—not the main server. 4. Architecture User ↓ CDN ↓ Origin Server 5. Key Takeaways ✅ Faster loading ✅ Lower latency ✅ Less load on origin server ✅ Better global performance 6. Question Why should a user in Japan fetch an image from the US when there's a copy nearby?

@LoyalDataEng All the best 👍

🚀 ANNOUNCING: AI Engineering Journey Over the next 10–12 weeks, I'm documenting my journey of becoming an AI Engineer—one episode at a time. Not by memorizing buzzwords. Not by copy-pasting tutorials. But by understanding how modern AI systems actually work. This series is for you if you're: • 🎓 A college student trying to break into AI • 💻 A Software/Data Engineer planning to transition into AI • 🤖 Curious about LLMs but don't know where to start • 📚 Tired of tutorials that either dive straight into math or only show demos Over 45 episodes, I'll break down and document my learning on: → LLMs & APIs → Prompt Engineering → RAG → Vector Databases → AI Agents → MCP → Evaluation → Deployment → Building production-ready AI applications I'm not claiming to know everything. I'm learning, building, making mistakes, and documenting everything in public. If you're on the same path, follow along. Episode 1 drops soon. 🚀 #AI #AIEngineering #LLM #BuildInPublic #GenerativeAI

@SidJain_80 That's what we are facing

@devXritesh Last line is hitting hard!! But reality

At 24 years, as an engineering graduate in AI era, I realized: • Good grades don't guarantee a job. • Rejections become normal. • Friends move ahead at different speeds. • Skills matter more than certificates. • Nobody owes you an opportunity. • The market isn't always fair. • Consistency is harder than talent. • Comparison steals your peace. • Building is better than waiting. • You have to keep believing when nobody else does. That's the reality many graduates never expect.

@ConsciousRide @Flipkart No because, I threw that product... So no proof ...

@devXritesh @Flipkart Did it get resolved later?

Very disappointing experience with @Flipkart. Today I ordered a Gokul milk packet through Flipkart Minutes and received a packet with the expiry date of the same day. After boiling, the milk curdled and became unusable. I immediately contacted customer support and shared all the details, but the complaint was simply closed without a refund or replacement. Customers should not be left bearing the cost of spoiled food products and poor support. Please look into this. #CustomerService Order ID: OD338067527730316500

@devfrom_hyd Ofcourse but need to learn how to solve problem with coding

Be honest In the era of Claude and ChatGPT, is coding still worth it?

@jahirsheikh8 Damn... Still AI detected

How I treat Claude Fable 5, 7 minutes before the essay deadline

@dharmvir_ Let do deal, don't raise this with QA guy

That moment when the QA guy discovers the dev planted the bug for job security.

@vivoplt Hehe... Good try 😅

ChatGPT is roaming the streets of Madrid🤣

One possibility is reading from a stale replica. The write went to the primary database, but the API read from a replica that hadn't caught up yet due to replication lag. Other possible causes: - Wrong database/tenant - Race condition - Data mapping bug I'd check the read path before blaming the database.

Backend Interview Question: The database contains: John124 The API returns: Mike567 No cache bug. How is this possible?