Pranav

Writing code and learning daily!!

[Pranav @Pranav3773]

It only took 16 minutes to spark the energy I needed for 10 hours of coding and learning.
@kunalbagaria 🙌💥 #web3

[Pranav @Pranav3773]

@Indujhac 😶😶

[Pranav @Pranav3773]

Wow, 7th sem and I’m sitting in a 9:30 AM class learning Parts of Speech 😭💀

[Pranav @Pranav3773]

LLMs are frozen in time + know nothing about your docs + hallucinate confidently when they don't know. RAG fixes that: Retrieve → Augment → Generate.

Naive chunking alone can tank faithfulness from 0.80 → 0.47. The pipeline is only as strong as its weakest link.

[Pranav @Pranav3773]

Agents sound complicated. They're not.
Just built one from scratch and it's an LLM, a few tools, and a while loop.
The model writes text, my code parses it, runs the tool, and feeds the result back. Repeat until done.

Demystified.
#AIAgents #Python #LearningInPublic

[Pranav @Pranav3773]

Built a mini-LLM from scratch on Shakespeare

Used:
→ RMSNorm + Pre-Norm
→ RoPE
→ GQA (8 Q / 2 KV heads)
→ SwiGLU
→ 4 Transformer blocks + weight tying
→ AdamW + gradient clipping
Loss: ~4.17 → Shakespeare-ish text.
Great lecture bhaiya @rishabh10x

[Pranav @Pranav3773]

LLM architecture since the original 2017 Transformer:

Turns out it's 4 upgrades:
→ LayerNorm → RMSNorm (simpler, faster)
→ ReLU → SwiGLU (learned gating)
→ Sinusoidal → RoPE (relative position via rotation)
→ Multi-Head → Grouped Query Attention (4-times less KV cache)

[Pranav @Pranav3773]

Every AI model — from a simple neural net to GPT-4 — is secretly doing ONE thing: matrix multiplication.
Scalars → Vectors → Matrices → Tensors (literally just stacking dimensions)
And the bonus part it feels too easy to understand when @rishabh10x is your mentor :)

[Pranav @Pranav3773]

@insanekrishnaa https://t.co/19G15ZL49b

[Pranav @Pranav3773]

My audio worked perfectly locally but stopped working after deployment.

The issue?
macOS is usually case-insensitive, while Linux servers are case-sensitive.
BTW you can review my portfolio though , it is not fully completed yet :) - https://t.co/19G15ZL49b

[Pranav @Pranav3773]

#5
It was the fun class many things revised first and then learned more concept:
~ Understand Self-Attention
~ Know Q, K, V
~ Know the Attention Formula
~ Understand Context-Aware Meaning
~ Understand KV Cache
~ Understand Multi-Head Attention and more....

[Rupesh Jagtap @rupesh248000]

@kirat_tw 😭😭 bhaiya me to DevOps padhne aya tha

[Pranav @Pranav3773]

#4
Initially it's little bit tough to understand but gradually all are digestible
Just because of @rishabh10x bhaiya🙌
: Text Tokenization Approaches & Their Trade-offs
: Practical Implications of Tokenization
: The Concept of Embeddings
: Semantic Vector Arithmetic... and more

[Pranav @Pranav3773]

#3
Great class, Learned a lot:
> Forward Pass: Data flows through layers to make a prediction.
> Calculate Loss: Measure exactly how wrong the prediction is.
> Backpropagation: Trace the error backward to assign "blame" to specific weights. And much more....
@rishabh10x 🫡

[Pranav @Pranav3773]

@ishaa_malhotra 👋

[Pranav @Pranav3773]

TakeAway from 2nd lecture:
~ Understanding Learning and Intelligence
~ Knowledge Distinction:
~ Evolution of AI Approaches
~ Natural Language Processing (NLP) Challenges
~ Word Embeddings (Words as Numbers)
~ Sequence Models vs. Transformers
How ChatGPT and LLMs Work,...much more

[Pranav @Pranav3773]

Did leetcode project, Key takeaways:

-how a worker swarm processes code submissions asynchronously.
-Backend → Queue → Workers → Execution → Polling → Result.
@kirat_tw 🙌

[Pranav @Pranav3773]

Just started AI/ML hope it goes well
@rishabh10x 🙌

[Pranav @Pranav3773]

Finally... it works. 😮‍💨

A couple of days ago, around 2 AM, I spent almost two hours trying to fix a bug, but nothing worked.

I finally gave up and went to sleep.
I was so frustrated that I was debugging the same bug in my dream.

Aur phir achanak se nind khul gyee💀

[Pranav @Pranav3773]

Here is the another virtual drive.

GD round!!!

[Pranav @Pranav3773]

@Harshita126062 Congrats 🎉