Tirthraj Mahajan

Age 22. Real life starts?

@kedar2sexy Not really

Rendering md files for docs has been a hassle for so long. Shitty web based editors, Very bad export to pdf experience and no shareable links at all Don’t get me started on No AI capabilities in these online md editors Is someone else facing this same issue?

And then there is Google. Total domination in all fields

At MTW 26. Majority of all presentations are about voice agents (multi-language)

@kedar2sexy Lekin tu to as frontend developer hire hua tha

Mere company me sirf Ui Ux karne ka reh gaya hu baki sab domains me to haat mar hi liya hai

I wrote a blog on DynamoDB internals - @tirthraj2004/dynamodb-internals-9e841b257704" target="_blank" rel="nofollow noopener">medium.com/@tirthraj2004/… Go check it out if you are interested!

The Minecraft to Computer Science pipeline needs to be studied

We will be here when you win every trophy possible at Chelsea football club Xabi ✨

AI made developers faster. It did not magically make them senior. I wrote about: - why output is becoming commoditized, - why judgment matters more than ever, - and what actually creates engineering depth in the AI era. Go check it out - @tirthraj2004/ai-made-you-faster-it-didnt-make-you-senior-ca0c2ce66f3b" target="_blank" rel="nofollow noopener">medium.com/@tirthraj2004/…

Why C++ still wins in finance.

One of the best videos I've seen recently. Definitely worth a watch!

Abstractions are important, they save time. But often, treating something as a complete blackbox doesn't help you grow more.

I didn't know how Python works underneath the hood and how GIL works. I implemented a basic version of it from scratch.

The best way to demystify things if to implement the goddamn thing itself. While it is not the "recommended" approach, I have always done this and it's always almost worked in my favor You don't have to implement it entirely, but it should fundamentally do the same thing

When the data model encodes the domain (like normalized SQL), bottom-up understanding works well. When the data model is optimized for access patterns (like DynamoDB), top-down understanding via APIs is more effective

As per my observation - Start with APIs (what does the system expose?) Peek into DB (how is it stored?) Map flows (how data moves) Then loop: API <-> DB <-> business logic

This is not really a hard and fast rule, though. The deeper principle is - Start from wherever the system’s 'design truth' lives. Some SQL systems (modern microservices) are API-first, and some NoSQL systems have implicit structure buried in data

However, having now worked with DynamoDB-based systems, their structure is optimized for access patterns, not relationships. If you start with the database schema, you'll misunderstand why the data is shaped the way it is. This is where top-down is more effective.

I have always worked with systems that use SQL as their underlying DB. Naturally, I go bottom up, looking at the database schema first, and then APIs, and then business logic. This is because traditional systems were designed database-first