Tina Sharma – Medium

Tina Sharma

Pinned

3 AI Learning Paths: Pick Yours (Access all my articles for free here)

A structured overview of machine learning, deep learning, neural networks, and large language models — everything you need to understand…

Apr 24

3 AI Learning Paths: Pick Yours (Access all my articles for free here)

Apr 24

Published in

Why Matrix Rank Is the Conceptual Link Between Embeddings, Bottleneck Layers, and LoRA

A plain-English guide to why a single number—the rank—decides how much information a neural network layer can carry and what happens when…

2d ago

The Matrix Property That Makes AI Forget

2d ago

Published in

Why Deep Learning Needs Matrices — Just Like Instagram Needs Filters

A precise guide to weight matrices, dimension changes, layer collapse, activation functions, and why W was never zero to begin with.

May 7

Why Deep Learning Needs Matrices — Just Like Instagram Needs Filters

May 7

Published in

If you think AI is like Google, just smarter, read this first.

Most people’s first mental model of AI is completely wrong. And that one wrong assumption is why AI keeps surprising, confusing, and…

May 7

If you think AI is like Google,
just smarter, read this first.

May 7

Published in

Most Machine Learning Candidates Don’t Know This Framework. The Top 10% Do.

How to answer any ML concept question — bias-variance as the worked example

May 6

A confident woman holding a glowing framework map stands at the front of a crowd facing a door labeled ML Interview, while others look confused behind her

May 6

Published in

In 2017, Google Stopped Fixing the Flaw. They Removed It.

How Attention Works: The Architecture That Broke the Vanishing Gradient

May 2

In 2017, Google Stopped Fixing the Flaw. They Removed It.

May 2

Published in

The Machine Learned to See. Then It Tried to Remember.

There was a mathematical wall hiding inside the most promising architecture in AI. Everyone could see it. Nobody could get around it. This…

Apr 30

The Machine Learned to See. Then It Tried to Remember.

Apr 30

Published in

Building Vector Search? Why FAISS Alone Isn’t Enough

What FAISS Does Well, Where It Stops, and When to Use a Vector Database Instead

Apr 28

Diagram showing the semantic search pipeline: input text data is converted into vectors using an embedding model, stored in a FAISS index, and queried to return the most similar results.

Apr 28

Published in

AI Product Managers: The 25 ML Terms That Actually Matter in Team Meetings

Most glossaries stop at definitions. This one tells you what goes wrong in your product when you get each term wrong — and what to ask…

Apr 28

AI Product Managers: The 25 ML Terms That Actually Matter in Team Meetings

Apr 28

Published in

There Are No Rules Inside a Trained AI

How neural networks learn: backpropagation, decision boundaries, and why AI doesn’t store rules

Apr 26

A wide illustration showing a glowing curved decision boundary dividing space into three regions labeled Output A, B, and C, with scattered data points and arrows indicating learning. A woman stands to the side observing the system, while a neural network diagram appears above.

Apr 26

Tina Sharma

Tina Sharma

Data Engineer. I love to explain the 'Why?' behind the AI concepts. https://linktr.ee/itinasharma Latest: https://whatsapp.com/channel/0029VbCuMxIGE56jhY8MBz15

Following

Help

Status

About

Careers

Press

Blog

Privacy

Rules

Terms

Text to speech