Webinar on "From Numbers to Meaning: Vectors, Embeddings, and How LLMs Understand Language"

#computer-society #vectors #learning #artificial-intelligence #industry-relations-committee #application #IEEE

IEEE Computer Society Hyderabad Chapter, in collaboration with the Industry Relations Committee of IEEE Hyderabad Section, invites you to an insightfulwebinar on:

“From Numbers to Meaning: Vectors, Embeddings, and How LLMs Understand Language”

Speaker: Mr. M K Pavan Kumar
Distinguished AI Architect | GenAI & RAG Expert, Equal AI

Date: 21 June 2026
Time: 10:00 AM IST
Registration Link: https://bit.ly/CS_Webinar3

Abstract:
This webinar explores how Large Language Models transform language into meaningful numerical representations using vectors and embeddings. Participants will gain an intuitive understanding of semantic similarity, word embeddings, positional encoding, and how modern LLMs use these concepts to understand context, meaning, and relationships in human language.

Join us to explore how vectors, embeddings, and LLMs power modern AI systems and language understanding.

A great opportunity for students, researchers, professionals, and AI enthusiasts to deepen their understanding of modern AI technologies.

#IEEE #IEEECS #IEEEHyderabadSection #ArtificialIntelligence #LLM #GenAI #MachineLearning #DataScience #Embeddings #RAG #AIWebinar #IEEEComputerSociety

This is high level agenda

Part 1 — Vectors

What is a vector? Intuition from geometry
Vectors in high-dimensional space
Why vectors are useful for representing meaning
Cosine similarity — measuring semantic closeness

Part 2 — From Words to Numbers

The core problem: how do machines read text?
One-hot encoding and its limitations
The idea of dense representations

Part 3 — Embeddings

What is an embedding?
Word2Vec intuition (King - Man + Woman = Queen)
GloVe — global co-occurrence
Contextual embeddings — why static embeddings fall short
BERT-style embeddings (awareness level)

Part 4 — Positional Embeddings

Why position matters in language
Sinusoidal positional encoding (original Transformer)
Learned positional embeddings
RoPE — intuition only, no heavy math

Part 5 — Embeddings inside LLMs

Token embeddings + positional embeddings combined
How attention uses embeddings to build meaning
Semantic search as a real-world application