AaradhyaSaxena

📄 WhatsApp Agent Design Doc: Meta Entity Integration

🎯 Objective

Enable WhatsApp-based meal planning and cook coordination via the WhatsApp Cloud API using a production-grade setup connected to Meta’s infrastructure.

---

🧱 Key Entities Overview

AaradhyaSaxena

Reading List

• The illustrated transformer: https://jalammar.github.io/illustrated-transformer/
  • pre-req: https://jalammar.github.io/visualizing-neural-machine-translation-mechanics-of-seq2seq-models-with-attention/
• Reasoning LLM's: A Visual Guide to Reasoning LLMs
• MoE: https://newsletter.maartengrootendorst.com/p/a-visual-guide-to-mixture-of-experts
• Quantization: https://newsletter.maartengrootendorst.com/p/a-visual-guide-to-quantization
• multi-head attention: Understanding and Coding Self-Attention, Multi-Head Attention, Cross-Attention, and Causal-Attention in LLMs
• Deepseek (multi head latent attention): https://medium.com/data-science/deepseek-v3-explained-1-multi-head-latent-attention-ed6bee2a67c4
• RoPE: https://karthick.ai/blog/2024/Rotatory-Position-Embedding-(RoPE)/

AaradhyaSaxena

RL

RL is the study of agents and how they learn by trial and error. It formalizes the idea that rewarding or punishing an agent for its behavior makes it more likely to repeat or forego that behavior in the future.

The main characters of RL are the agent and the environment. The environment is the world that the agent lives in and interacts with. At every step of interaction, the agent sees a (possibly partial) observation of the state of the world, and then decides on an action to take. The environment changes when the agent acts on it, but may also change on its own.

The agent also perceives a reward signal from the environment, a number that tells it how good or bad the current world state is. The goal of the agent is to maximize its cumulative reward, called return. Reinforcement learning methods are ways that the agent can learn behaviors to achieve its goal.

RL terminology:

• states and observations,

AaradhyaSaxena

For a task like PII (Personally Identifiable Information) data redaction, the training objectives and methodology should focus on identifying and masking sensitive information in text. Below are the most relevant training objectives and approaches for your use case:

---

1. Training Objectives

a. Token Classification (Named Entity Recognition - NER)

• Purpose: Teach the model to identify spans of text corresponding to PII entities (e.g., names, addresses, SSNs, credit card numbers).
• Method:
  • Frame the problem as a token classification task with labels such as O (non-PII), B-PII (beginning of PII entity), and I-PII (inside a PII entity).
• Fine-tune a pretrained model (e.g., BERT, DistilBERT, or Flan-T5) on labeled data.

AaradhyaSaxena

References

LSTM

• https://colah.github.io/posts/2015-08-Understanding-LSTMs/
• (unfinished) Attention and Augmented RNN's - https://distill.pub/2016/augmented-rnns/
• (unfinished) Visualizing memorization in RNNs - https://distill.pub/2019/memorization-in-rnns/

Unread

Circuits

AaradhyaSaxena

# Qdrant

Vector databases are a relatively new way for interacting with abstract data representations derived from opaque machine learning models such as deep learning architectures. These representations are often called vectors or embeddings and they are a compressed version of the data used to train a machine learning model to accomplish a task like sentiment analysis, speech recognition, object detection, and many others.

These new databases shine in many applications like semantic search and recommendation systems, and here, we’ll learn about one of the most popular and fastest growing vector databases in the market, Qdrant.

Concepts

1. Collection

AaradhyaSaxena

These are my notes from the book "Introduction to information retrieval"

1. Intro to IR
2. Term Vocabulary
3. Tolerant Retrieval
4. Index Construction
5. Index Compression
6. Scoring
7. Compute Scoring
8. Evaluation of Information Retrieval

AaradhyaSaxena

System Design

System design is about — Architecture + Data + Applications

Architecture means how are you going to put the different functioning blocks of a system together and make them seamlessly work with each other after taking into account all the nodal points where a sub-system can fail/stop working.

Data means what is the input, what are the data processing blocks, how to store petabytes of data and most importantly how to process it and give the desired/required output.

AaradhyaSaxena

Design Patterns

Design patterns can be classified into three main categories:

Creational Patterns:

Creational design patterns relate to how objects are constructed from classes. Examples of creational patterns include Factory Method, Abstract Factory, Builder, Prototype, and Singleton.

Structural Patterns:

AaradhyaSaxena

Spring Basics

The highlight of Spring Boot is its auto-configuration feature whereby it automatically includes all the dependencies of a project based on property files and JAR classpaths. Spring Boot is basically the Spring framework along with embedded servers. Spring boot removes the need for XML configuration.

Decoupling Components

public interface Filter {
    public String[] 
    getRecommendations(String movie);
}