Most developers are learning how to use AI or Calling/Consume LLM API
But the real skill is learning how to build reliable AI systems.
A production AI app needs more than an LLM:
→ Prompt engineering
→ RAG & embeddings
→ Vector databases
→ Tool calling
→ Agent workflows
→ Evaluation
→ Observability
→ Cost & latency optimization
→ Security & guardrails
Calling an LLM API is easy.
Building an AI system that users can trust is the real engineering challenge.
That’s where AI Engineering gets interesting.
#AI #ArtificialIntelligence #AIEngineering #LLM #GenerativeAI #RAG #MachineLearning #SoftwareEngineering #Developers #Startup
One thing I learned about State Management
Not every piece of state belongs in Redux.
Before choosing a state management solution, first ask:
Who needs this data?
How long should it live?
Does it need to be shared?
Is it server data or UI state?
The architecture starts with these questions.
Think about a typical e-commerce app 🛒
• Button loading → Local state
• Modal open/close → Local/UI state
• Filters → URL state
• Logged-in user → Global state
• Products from API → Server state
• Shopping cart → Shared/global state
Different problems → different state solutions.
A common mistake is putting everything into Redux.
But server data like:
Users, Products, Orders, Posts
often needs:
• Caching
• Refetching
• Synchronization
• Loading/error states
• Background updates
That's where tools like TanStack Query can be more appropriate.
Redux ≠ Server State Manager for every use case.
A simple mental model I use:
🟢 Local state → component
🔵 URL state → browser URL
🟠 Server state → TanStack Query
🟣 Global client state → Redux/Zustand
Good state management isn't about choosing the "best" library.
It's about putting each state in the right place.
Thanks
#React #Angular #UI #SystemDesign #WebDevelopment #Frontend #Startup
Today I learned when to use different rendering approaches:
🔹 CSR — Client-Side Rendering
Best for highly interactive apps like dashboards, admin panels, or internal tools where SEO isn't a priority.
🔹 SSR — Server-Side Rendering
Useful for dynamic, SEO-friendly pages like product pages, news sites, or personalized content that needs fresh data.
🔹 SSG — Static Site Generation
Great for content that doesn't change often — blogs, documentation, landing pages, and marketing websites.
🔹 ISR — Incremental Static Regeneration
Useful when content changes occasionally, like e-commerce product pages or news articles, but we still want the benefits of static pages.
🔹 Streaming SSR
Helpful when some parts of a page are fast and others are slow. The server can send the ready parts first instead of making the user wait for the entire page.
For startups or any organisation choosing the right approach can make a big difference in performance, scalability, infrastructure cost, SEO, and user experience.
The key lesson for me:
Don't choose a rendering strategy because it's popular. Choose it based on the content and the problem you're solving.
Thanks
#Frontend #WebDevelopment #SystemDesign #Startup #React #JavaScript
Building a startup? Don't start with microservices. 🚫
Start with a simple architecture that can evolve.
A practical path:
1️⃣ Start with a Monolith
→ Faster development
→ Easier debugging
→ Lower infrastructure cost
2️⃣ Add PostgreSQL
→ Proper indexes
→ Connection pooling
→ Read replicas when needed
3️⃣ Add Redis
→ Cache frequently accessed data
→ Reduce DB load
4️⃣ Add a Queue
→ Move heavy tasks to background workers
→ Emails, notifications, reports, etc.
5️⃣ Scale Horizontally
→ Load Balancer
→ Multiple application instances
→ Stateless APIs
6️⃣ Split Services only when necessary
→ Clear business boundaries
→ Independent scaling
→ Independent deployments
The goal of startup architecture isn't to build the most complex system.
#SystemDesign #SoftwareEngineering #Startups #Architecture #Backend
Want to build better frontend designs using AI coding agents? 👀
I recently came across these 2 interesting open-source projects:
🔹 Awesome DESIGN.md
https://t.co/ykCo1dT4il
🔹 Web Design
https://t.co/JQQ0mswnBF
The interesting part is that instead of simply asking an AI agent to “build a UI”, you can give it design context, patterns and references first.
That can help AI coding agents generate more consistent and better-looking frontend designs.
AI coding is not just about generating code anymore.
Design context + AI agent + code = much better frontend development. 🚀
#AI #AICoding #Frontend #WebDevelopment #React #Angular #OpenSource #Coding #Startup
Working with a shared library and a client application?
A typical workflow looks like this:
1️⃣ Develop the library
2️⃣ Build the library → "npm run build"
3️⃣ Generate the package → "dist/"
4️⃣ Publish it to nexus which I use https://t.co/HtjFfpQoHk
5️⃣ Client repo installs the required version
6️⃣ Client repo imports and uses the library.
For local development, you can also use:
"npm link"
The important part is keeping the library contract stable — inputs, outputs, APIs and versioning.
Library development isn't just writing reusable code.
It's designing a contract that other repositories can safely depend on.
#SoftwareEngineering #Frontend #Libraries #NPM #Developer #StartupDevelopment #Startup
📚 Today’s learning:
Sometimes LLMs generate outdated code or suggest APIs that no longer exist.
Why? 🤔
LLMs have a knowledge cutoff, so their built-in knowledge may not reflect the latest version of a library.
💡 A useful solution: Context7
Context7 is an MCP server that fetches up-to-date, version-specific documentation and provides it as context to the LLM.
So instead of guessing from outdated knowledge, the LLM can work with the latest docs for the exact version you're using.
🔗 https://t.co/vdEqm5qpx5
#AI #LLM #MCP #Context7 #DeveloperTools #Coding