manish

[manish @manishDebugs]

Selected for Amazon ML Summer School 2026🥳

[manish @manishDebugs]

#RCBvsGT thin air, flat deck, fast oufield, dew, if best bowling side conceded 250+, will we able to defend this. We need some early wickets🫠

[manish @manishDebugs]

#RCBvsGT @rrjjt_01 O Captain! My Captain!

[manish @manishDebugs]

@Spicy342 https://t.co/2O7YFQGFFa

[manish @manishDebugs]

Finally, i got my portfolio website done😁;)

[manish @manishDebugs]

@GuptaSayujya Happy birthday

[manish @manishDebugs]

@striver_79 Hey @striver_79 , I totally messed up the deadline, thought it was midnight, but your tweet just set me straight. Any chance I can still fill out the form? I know it's silly of me not to have read it properly, but I promise I won't add any commits, I even deployed the project.

[manish @manishDebugs]

@NickK2305 1. https://t.co/nhtEw6Ks0N 2.https://t.co/pw7Rz0b75J

[manish @manishDebugs]

@Spicy342 Aadi bhaiii

[manish @manishDebugs]

@DevByLakshya Interested

[manish @manishDebugs]

what is required qualification or skill for this role??! #rockstargames

[manish @manishDebugs]

@Beast__07_

[manish @manishDebugs]

>Duplicate request collapsing: 100,000 users ask for the exact same thing at once? System fires ONE request to backend. Everyone else just reuses the same response.

[manish @manishDebugs]

When match ends more than 10 MILLION people instantly quit and smash the homepage at the EXACT same second.
Each homepage hits 6 APIs (recommendations, trending, continue watching, ad,etc).
That’s 60,000,000 requests in one heartbeat. Ever wondered how @JioHotstar deals it:

[manish @manishDebugs]

>Personalization? Not computed live. Heavy ML recommendations run in batch pipelines hours earlier, stored in blazing-fast Redis. API becomes a simple lookup. Zero heavy lifting during the storm.

[manish @manishDebugs]

>Step autoscaling reacts too slowly — by the time new servers launch, the traffic spike has already hit.
>Even availability zone placement can slow things down because the cloud keeps retrying zones before launching instances.

[manish @manishDebugs]

In this video, Gaurav(@OyeHooye ) explains why @JioHotstar had to build their own autoscaling policies instead of relying on default autoscaling from Amazon Web Services.

At Hotstar scale, things break in unexpected ways:

[manish @manishDebugs]

this video of @OyeHooye should be a must-watch for every backend engineer building at scale.

[manish @manishDebugs]

>You start getting “insufficient capacity” errors when trying to spin up thousands of instances — AWS basically says “we don’t have that many right now.”
>Using a single instance type in an autoscaling group? If that instance runs out in the region, your scaling just stops.