Jiya Agrawal

Full Stack Developer (MERN) | Building scalable web apps & sharing what I build & break | GitHub + Blogs as proof | Simplicity is my superpower | GLAU '27

[Jiya Agrawal @JiyaAgr]

Day 57 - Today, I learned the basics of NumPy, focusing on:

ndarray - understanding NumPy's core array object.

Array Creation - learning different ways to create NumPy arrays.

[Jiya Agrawal @JiyaAgr]

Day 56 : NumPy & Pandas for AI/ML Engineers

Python lists offer flexibility. NumPy arrays bring efficient numerical computing through vectorization.

Learning the foundations behind AI, one concept at a time.

More tomorrow.

#Python #NumPy

[Jiya Agrawal @JiyaAgr]

Day 55
Accuracy: Overall correct predictions.
Precision@K: Relevant results among the top K.
Recall@K: Relevant results retrieved in the top K.
MRR: Rank of the first relevant result.
MAP: Average precision across queries.
NDCG: Quality of ranked results.

[Jiya Agrawal @JiyaAgr]

Day 54 - Today's ML learning: Precision , Recall & F1 Score — measuring classification model performance.

#MachineLearning #AI

[Jiya Agrawal @JiyaAgr]

Day 53

Learned something fascinating about Product Quantization (PQ) today

8 subvectors × 256 centroids = (2^{64}) possible combinations using just 8 bytes per vector.

Small codebooks, massive possibilities, efficient vector search.

#AI #MachineLearning

[Jiya Agrawal @JiyaAgr]

Day 52 - Learning Probability Distributions for AI

Embeddings → Represent data as vectors
Gaussian Distribution → Model data patterns
Distance Concentration → Distances become more similar in high dimensions
Real Use → Faster vector search & AI retrieval
#AIEngineering

[Jiya Agrawal @JiyaAgr]

Day 51 - Today I Learned: Probability Distributions in ANN

Distance concentration in high dimensions
Gaussian distributions and embeddings
Recall@K and search accuracy
How data distribution impacts ANN performance

#MachineLearning #AI

[Jiya Agrawal @JiyaAgr]

Day 50 - Hashing

LSH → Find similar vectors faster
MinHash & SimHash → Detect similar sets and documents
Hamming Distance → Compare binary hash codes
Real Use → Similarity search & duplicate detection

#AIEngineering #MachineLearning

[Jiya Agrawal @JiyaAgr]

Day 49 - Today, I learned about Random Projections in Machine Learning.

Dimensionality reduction

Johnson–Lindenstrauss Lemma

Random Projection vs. PCA

Faster similarity search and document retrieval
#MachineLearning #AI

[Jiya Agrawal @JiyaAgr]

Day 48 - Learned about Space Partitioning for faster Nearest Neighbor Search

KD-Tree: divide a city into areas and search nearby ones.
Ball Tree: group nearby places into zones.
Pruning: skip regions that are clearly too far.
High dimensions: these methods become less effective

[Jiya Agrawal @JiyaAgr]

Day 47
Explored Minkowski Distance, Jaccard Similarity & Hamming Distance and their real-world use in KNN, recommendations, similarity search & duplicate detection.

[Jiya Agrawal @JiyaAgr]

Day 46 - Graph Theory & Combinatorics

Graphs → Model connections like users & networks
Combinatorics → Count possible choices
Permutations → Arrange things where order matters
Combinations → Select things where order doesn’t matter

#AIEngineering

[Jiya Agrawal @JiyaAgr]

Day 45 - Functions in AI

Functions → Map inputs to outputs
Injective / Surjective / Bijective → Different types of mappings
Composition → Connect multiple functions
Real Use: Neural network layers are functions that transform data.

[Jiya Agrawal @JiyaAgr]

Day 44- Sets & Relations

Sets → Organize data into groups
Set Operations → Find common or unique data
Relations → Connect different data sets
Real Use → Databases, users, orders & products
Example → Users <-> Orders

#AIEngineering #MachineLearning

[Jiya Agrawal @JiyaAgr]

Day 43 - Multivariable Calculus for AI

Partial Derivatives - Understand individual variable impact
Chain Rule -Track changes through multiple layers
Gradient -Find the direction to improve
Jacobian -Track multiple inputs & outputs
Hessian -Understand the shape of optimization

[Jiya Agrawal @JiyaAgr]

Day 42- Limits & Derivatives

Limits → How values behave near a point
Derivatives → Measure how things change
ML → Gradients help reduce model loss
Real Use → Neural networks & optimization

[Jiya Agrawal @JiyaAgr]

Day 41 - Learning Rate & Minima

Learning Rate → Controls the step size
Small Rate → Slower but safer
Large Rate → Faster but may overshoot
Local Minimum → Best point in a local region
Global Minimum → Lowest point overall

#AIEngineering #MachineLearning

[Jiya Agrawal @JiyaAgr]

Day 40 - Gradient Descent → Finding the Best Solution

Gradient Descent → Move toward lower loss
Learning Rate → Controls the step size
SGD → Updates using smaller data samples
Real Use → Training ML models

#AIEngineering #MachineLearning

[Jiya Agrawal @JiyaAgr]

Day 39

Loss Function → Measures how wrong a model’s predictions are.

Convex vs Non-Convex Functions → Describes the shape of a function and how easy it is to optimize.

#AIEngineering #MachineLearning

[Jiya Agrawal @JiyaAgr]

Day 38 - Understanding Objective Functions

Objective Function → Defines what to optimize
Maximization → Increase revenue or accuracy
Minimization → Reduce cost or error
Constraints → Define what’s allowed
ML → Training optimizes the objective

#AIEngineering