---
title: "Ayush Pandit (@ayush27p) | Hashnode"
description: "Curious about Scalable ,Distributed Systems and AI. Read the latest articles by Ayush Pandit on Hashnode."
url: "https://hashnode.com/@ayush27p"
---

@ayush27p

# Ayush Pandit

@ayush27p · Bangalore, India · Joined June 2026

Curious about Scalable ,Distributed Systems and AI

### [Understanding the Limitations of Pubsub Systems](https://ayush-pandit.hashnode.dev/understanding-the-limitations-of-pubsub-systems?utm_source=hashnode&utm_medium=feed)

Paper: Understanding the Limitations of Pubsub Systems Authors: Atul Adya, Phil Bogle, Colin Meek (Databricks) Venue: HotOS 2025 If you've ever built anything with Kafka, GCP Pub/Sub, RabbitMQ, or AWS

13 min read · [ayush-pandit.hashnode.dev](https://ayush-pandit.hashnode.dev/understanding-the-limitations-of-pubsub-systems?utm_source=hashnode&utm_medium=feed)

[Aug 23](https://hashnode.com/posts/understanding-the-limitations-of-pubsub-systems/6a8b43f5b941c03892755ea7)

### [Write-Through vs Write-Behind Caching: How High-Performance Systems Handle Writes](https://ayush-pandit.hashnode.dev/write-through-vs-write-behind-caching-how-high-performance-systems-handle-writes?utm_source=hashnode&utm_medium=feed)

1\. Why Do We Need Caching for Writes? Most caching conversations start and end with reads: "the database is slow, put Redis in front of it." But in any system with real traffic, writes become the bott

22 min read · [ayush-pandit.hashnode.dev](https://ayush-pandit.hashnode.dev/write-through-vs-write-behind-caching-how-high-performance-systems-handle-writes?utm_source=hashnode&utm_medium=feed)

[Aug 19](https://hashnode.com/posts/write-through-vs-write-behind-caching-how-high-performance-systems-handle-writes/6a853cfb8928e0844ae89da1)

### [Replication Is Expensive. Here's How S3 Gets the Same Durability for Half the Cost](https://ayush-pandit.hashnode.dev/replication-is-expensive-here-s-how-s3-gets-the-same-durability-for-half-the-cost?utm_source=hashnode&utm_medium=feed)

A deep dive from first principles to Reed-Solomon math to production architecture Why this matters If you've ever wondered how S3 promises 99.999999999% durability (eleven nines lose 1 object out of

14 min read · [ayush-pandit.hashnode.dev](https://ayush-pandit.hashnode.dev/replication-is-expensive-here-s-how-s3-gets-the-same-durability-for-half-the-cost?utm_source=hashnode&utm_medium=feed)

[Jul 1](https://hashnode.com/posts/replication-is-expensive-here-s-how-s3-gets-the-same-durability-for-half-the-cost/6a456277c55140ec84efa96e)

### [How Google Lens Works: From Uploaded Image to Search Results](https://ayush-pandit.hashnode.dev/how-google-lens-works-from-uploaded-image-to-search-results?utm_source=hashnode&utm_medium=feed)

1\. Introduction: why reverse image search is a hard engineering problem Here's the trap juniors fall into with this topic: they assume the hard part is "the AI." It isn't. The model that turns an imag

16 min read · [ayush-pandit.hashnode.dev](https://ayush-pandit.hashnode.dev/how-google-lens-works-from-uploaded-image-to-search-results?utm_source=hashnode&utm_medium=feed)

[Jun 28](https://hashnode.com/posts/how-google-lens-works-from-uploaded-image-to-search-results/6a416a12278391c7a9004be7)

### [MongoDB Change Streams Internals From Database Writes to Real Time Systems](https://ayush-pandit.hashnode.dev/mongodb-change-streams-internals-from-database-writes-to-real-time-systems?utm_source=hashnode&utm_medium=feed)

If you ask ten backend engineers how MongoDB Change Streams work, most will answer something like: "You call watch() and MongoDB notifies you whenever a document changes." That answer is technically

9 min read · [ayush-pandit.hashnode.dev](https://ayush-pandit.hashnode.dev/mongodb-change-streams-internals-from-database-writes-to-real-time-systems?utm_source=hashnode&utm_medium=feed)

[Jun 28](https://hashnode.com/posts/mongodb-change-streams-internals-from-database-writes-to-real-time-systems/6a411b90defcef0e3494d419)
