Google is on a downfall and might become a legacy company as compared to the new high tech cool company.
•Google cloud sucks as compared to Azure and AWS. •Pixel phones have heating and network issues.
•Gemini s*cks, open source models will be way better.
Deep Learning became popular because of huge data, powerful GPUs, and better algorithms.
More data + more compute = smarter models.
That’s the real reason behind today’s AI boom. 🚀
Average Word2Vec
Instead of one-hot or TF-IDF,
each word has a dense vector.
Sentence representation =
average of all word vectors.
Simple ✔️
Keeps semantic meaning ✔️
But loses word order ❌
Good baseline, not the final answer.
#Word2Vec#NLP#GenAIJourney#LearningInPublic
Day 10 | TF-IDF
Learned how NLP measures
word importance in text.
TF → frequency in a document
IDF → down-weights common words
Pros: intuitive, captures importance
Cons: sparse vectors, OOV issue
Small concept, big impact.
#Day10#TFIDF#NLP#GenAIJourney#LearningInPublic
Day 9 – Progress Check
What changed in 9 days:
Fear ↓
Clarity ↑
Curiosity ↑↑
Still a beginner.
Still confused sometimes.
But no longer stuck.
Consistency > Motivation.
Onwards 🚀
#Day9#LearningInPublic#GenAIJourney
Day 6 – Backend for AI
Explored:
Flask
FastAPI
How ML models get exposed as APIs
AI is useless if you can’t serve it.
Learning how “model → product” actually happens.
#Day6#FastAPI#MLDeployment
Day 4 – One Hot Encoding
Learned:
What is One Hot Encoding
Advantages & limitations
Simple but powerful idea:
More features ≠ better always.
Vocabulary size matters more than I thought.
#Day4#MachineLearning#NLP