Jaspinder Singh

I teach hardcore DSA, AI and ML.

[Jaspinder Singh @JaspinderS30]

Word Tokens —> [“hello” , “world”]

Subword Tokens —> [“he, “llo”, “wor”, “ld”]

BPE - Byte Pair Encoding (Subword Tokenization) —> Used in GPT

[Jaspinder Singh @JaspinderS30]

Protests are done.

Let's learn AI ML now and build some LLMs.

[Jaspinder Singh @JaspinderS30]

More at https://t.co/hbJtNIppUb. Only for hardcore AI ML understanding

[Jaspinder Singh @JaspinderS30]

You built a transformer ..great.

But before you start training your LLMs from scratch.

Follow this to avoid unstability and making costly computational mistakes

[Jaspinder Singh @JaspinderS30]

5. Batch Sizing

[Jaspinder Singh @JaspinderS30]

“Learn AI” is the new “Learn to Code”

[Jaspinder Singh @JaspinderS30]

Or do they simply learn to use frameworks without understanding what's happening under the hood?

Curious to hear different teaching approaches.

#LLM #GenAI #MachineLearning #DeepLearning #AIEducation

[Jaspinder Singh @JaspinderS30]

Genuine question for GenAI course instructors.

If you start directly with LLMs and GenAI, how do your students truly understand:

• Self-Attention weight calculations?
• Q, K, V tensor transformations?
• 4D tensor broadcasting across batches and heads?
• Why masking works?

[Jaspinder Singh @JaspinderS30]

Multi-head self-attention

[Jaspinder Singh @JaspinderS30]

An {S30} alumn is looking of ML Engineer for Atlassian. If interested pitch yourself in 1-2 sentences only in reply to this tweet

#MLEngineer #MachineLearning #Hiring #Atlassian #AIJobs

[Jaspinder Singh @JaspinderS30]

Most students skip understanding gradient derivations == Most students are confused about gradients == Most students dont understand how AI/ML works

[Jaspinder Singh @JaspinderS30]

Interview Prep?

Later ….Later …..Later …Later

Layoffs Happen

OH MY GOD…OH MY GOD…I have just 60 days now to prep

Then …

Disaster

[Jaspinder Singh @JaspinderS30]

Announcing offers of the week

Google AI ++

Meta AI ++

Open AI ++

through https://t.co/hbJtNIppUb platform.

We have added FAANG video explanations to most labs now

[Jaspinder Singh @JaspinderS30]

The AI stack evolves fast, but the fundamentals remain surprisingly durable.

[Jaspinder Singh @JaspinderS30]

Every week, Twitter fills up with discussions on the latest architectures, reasoning models, and research papers.

[Jaspinder Singh @JaspinderS30]

Meanwhile, many of the world's top AI companies still ask questions like: Why does regularization help? Why ReLU over sigmoid? Why cross-entropy instead of MSE?