Medium

MediumShubham Patel – Medium

Shubham Patel

he

Shubham Patel

·

23h ago

Redis for System Design Interviews: Why It’s Fast, When to Use It, and Every Question You’ll Be…

A beginner-friendly guide to Redis: what it is, why one thread can serve a million requests a second, how to use it as a cache without…

Redis for System Design Interviews: Why It’s Fast, When to Use It, and Every Question You’ll Be…

Shubham Patel

·

2d ago

Design Twitch: The System Design Interview, Explained Simply

Ingest, real-time transcoding, a CDN that never gets a head start, and a chat room where 200,000 people talk at once. A step-by-step…

Design Twitch: The System Design Interview, Explained Simply

Shubham Patel

·

Nov 4

Open Another Till: Kubernetes HPA, Explained Simply

How the Horizontal Pod Autoscaler decides when to add or remove Pods, why CPU requests matter so much, and the brakes that stop it from…

Open Another Till: Kubernetes HPA, Explained Simply

Shubham Patel

·

Nov 2

The Pod You Wrote Is Not the Pod That Runs: Kubernetes Admission Webhooks, Explained Simply

Mutating webhooks tag it, validating webhooks scan it. A friendly tour with a working Go webhook, real error messages from a live cluster…

The Pod You Wrote Is Not the Pod That Runs: Kubernetes Admission Webhooks, Explained Simply

Shubham Patel

·

Oct 31

Ingress or Gateway API? Picking Your Cluster’s Front Door on AWS ALB

Four little words — plan, build, route, check — and the ALB stops being a black box

Ingress or Gateway API? Picking Your Cluster’s Front Door on AWS ALB

In

GoPenAI

by

Shubham Patel

·

Oct 30

LLM: Prompt, RAG, or Fine-Tune?

The most common GenAI architecture mistake is fine-tuning when you needed retrieval. Here’s how to tell which lever a problem needs.

LLM: Prompt, RAG, or Fine-Tune?

Shubham Patel

·

Oct 29

Why the Answer Took Five Seconds

The first word arrived in 400 ms. The rest took five seconds. Those are two different problems with two different fixes.

Why the Answer Took Five Seconds

Shubham Patel

·

Oct 28

Backpressure for System Design Interviews: Why Fast Systems Collapse, How Good Ones Say “Slow…

A beginner-friendly guide, from a busy Dosa restaurant to Kafka, TCP and gRPC, so you can explain backpressure to anyone, including your…

Backpressure for System Design Interviews: Why Fast Systems Collapse, How Good Ones Say “Slow…

In

AEO Soup

by

Shubham Patel

·

Oct 27

LLM: Meaning as Coordinates

How software finds “Refund policy” when you typed “can I get my money back,” and what that kind of search can’t tell you

LLM: Meaning as Coordinates

Shubham Patel

·

Oct 27

Headless Services in Kubernetes: When Every Pod Needs Its Own Name

Why Kafka can’t live behind one phone number, and four words that make headless Services click: None, List, Name, Card

Headless Services in Kubernetes: When Every Pod Needs Its Own Name

Shubham Patel

Shubham Patelhe

Senior Software Engineer and Technical content creator passionate about making complex technical concepts accessible to developers at all levels.

Following

Help

Status

About

Careers

Press

Blog

Store

Privacy

Rules

Terms

Text to speech