Data Science, Database, AI Startups and Domain Learning's (Video-Image-Text-Data-Database)

October 15, 2023

CNN Learning One pagers

Product and Example

https://tangoeye.ai/
Retail solutions built on
Age Detection Models
Gender Detection Models
Face Detection
Re-identification

Models for Training - tensorflow hub

How CNN works - Visualizer

How Features are Learned - 10 class classification

But what is a neural network

The Complete Mathematics of Neural Networks and Deep Learning

Learning Procedure

Step 1: Take a batch of training data and perform forward propagation to compute the loss.
Step 2: Backpropagate the loss to get the gradient of the loss with respect to each weight.
Step 3: Use the gradients to update the weights of the network.

Backprop summary

Chain rule derivate
The procedure repeatedly adjusts the weights of the connections in the network so as to minimize a measure of the difference between actual output and desired output
Ability to create new distinguishing features
The aim is to find the set of weights that ensure that for each input vector the output vector produced by the network is same as the desired output vector
The drawback in learning procedure is that the error surface may contain local minima so that gradient descent is not guaranteed to find a global minimum

Activation Functions

Introduce non-linearity into a model
We need non-linearity, to capture more complex features and model more complex variations that simple linear models can not capture.
neural networks use non-linear activation functions, which can help the network learn complex data, compute and learn
Signmoid, Tanh, Relu

Designing CNN

The first rule of thumb is that you should not try to design your own architecture from scratch
If you are working on generic problem, it never hurts to start with ResNet-50. If you are building a mobile-based visual application where there is limited computation resources, try MobileNets

Keep Learning!!!

October 12, 2023

Forecasting Setup MVP to Large Scale Deployment

Phase I (Initial use cases)

DVC - Data Versioning and Control
Databases - Sales / Historical data
External Marketing Campaign data
Feature Engineering - Database / Feature Store
Model Building - ML / DL Algos
Model Experiment Tracking - MLflow
Drift / Monitoring - Evidently AI
Model Deployment - API / Serverless function
Actuals Tracking Loading - Database
Model Scheduling / ETL Scheduling - Jobs based on AWS / GCP / Custom scheduler cron jobs
I prefer dockerizing it and deploy it in GCP APP engine like a Nocode / Low code approach for the first few models
Custom Reporting for trends/patterns making it more accessible/relatable to business
Reporting for Past - Present - Future. Making predictions relatable
Explainable AI to compare predictions vs actuals to interpret cause - -reason in a more Non-ML Approach

Phase II (Serving large models > 30)

Kubernetes based Platform
Once the platform is deployed you can leverage out of box images / notebook / pipeline / monitoring options available
Kubeflow Pipelines

Kubernetes as a Service: GKE vs. AKS vs. EKS

Build a Pipeline

Keep Exploring!!!

October 11, 2023

Chatbot Lessons

Interesting Read

GenAI Customer Care Project Now Saving $4M/yr for Car Insurance Firm, Jerry

Key Lessons

Leverage Large Language Models (LLMs) to create a chatbot
Open AI GPT-4 for complex queries
Open AI GPT-3.5 for initial sorting
Dataset - Messages are captured from chat and SMS through Twilio and stored on Jerry's servers
Routing Agent - Route based on requests - "Payments," "Policy," and "Opt-out" agents
Webhook / API - Handlebars to insert data from their database into the chatbot's responses

Key Learning's

Prompt engineering
Rapid iteration
Investment in testing
Version control

Webhooks in Python with Flask

Webhook

How to Create a Webhook in Python

Python Flask Webhook Receiver

Drawback of LLM

Exposes your corporate data to the provider of your LLM
LLMs have been shown to suffer from “hallucinations

Quick Lessons

FAQ from Cache
VectorDB for queries where answers can be located in docs
Intent recognition and call APIs based on OrderNumber#
Knowledge Graph if we have some meta data loaded
If all attempt fails LLM answer
Add guard rails wherever possible

Ref - Link

Dialogflow Notes Link

Keep Exploring!!!!

Vision Latest Notes - Text-to-Image Generation

Alignment in Text-to-Image Generation

Key components

Controllable Generation
Editing
Better following prompts
Customization

Techniques

GAN
Auto-Regressive
Diffusion
Non AR Transformer

One Liners

GAN - Learn to Fake it until it becomes Real
AR - Image to patches, Patches mapped to indexes - Tokens, Prediction - Set of tokens one by one
Non-AR Transformer - Schedule policy to generate tokens at Each step
Diffusion - Random noise in each step, Subtract noise to end up with the required semantic quality

Diffusion Overview

GLIDE

DALL-E-2

Image Super-Resolution via Iterative Refinement

Keep Exploring!!!

Evaluating LLM Models

Good checklist on Evaluating LLM Models

LLM Type - Open source / Proprietary / Datasets
Deployment Options - Private cloud or API model
Infra needs - Self-hosting
Retrieval Augmentation - Support for RAG
Scalability - Performance of large RAG datasets
Hallucinations - Handling Hallucinations
Benchmark compared to other competitive models
Legal Compliance - IP / ownership of prompts / PII
Output Compliance - Bias / Toxicity
Output filtering / Content filters

Keep Exploring!!!

Sales Forecast - Exogenous vs Endogenous Variables

Sometimes jargon occupies too much of concepts, Being able to explain it in business terms is key to being able to relate to it.

Now we have data drift, and model drift. 13 years ago when we implemented forecast when the actual and prediction varied we ask the store manager

Reason for it
Does actual data make sense for future training or use forecasted data

Terms Exogenous / Endogenous can be put in more business-friendly Terms :)

Exogenous - External cause

1. Market Trends: This refers to the overall direction in which the market that the product or service is operating in is headed. It includes factors such as consumer behavior, buying trends, or preferences which can affect the sales forecast significantly.

2. Economic Conditions: This includes factors like inflation rates, unemployment rates, gross domestic product (GDP), etc. These conditions can influence consumer spending and thus directly affect the sales forecast.

3. Competition: The actions (like pricing strategies, marketing campaigns, product launches) of competitors in the market can also significantly impact the sales of a product or service. Hence, this is an important exogenous variable in sales forecast.

Endogenous - Internal cause

1. Past Sales Data: The historical sales performance is a key factor in predicting future sales. Based on historical data, businesses can form patterns and trends that help in forecasting future sales.

2. Pricing: The price of a product or service plays a crucial role in determining its demand. Changes in price, due to factors like discounts, offers, etc., can significantly influence sales.

3. Advertising and Promotions: The level of advertising and promotional activities can impact the demand of a product or service. Businesses can increase sales by intensifying their marketing efforts. Therefore, the effectiveness of advertising and promotional activities is directly related to the sales forecast.

Keep Exploring!!!