AWS SaaS Blueprint

Interesting session on building SaaS Apps. Include different aspects hosting, keeping multiple versions, Isolating resources, sharing

• Categorize patterns / Building Blocks
• Manage and Operate Business
• Tenant aware, Tier aware metrics

• Flavors of Isolation
• Parition workloads
• Partition data

• Domain / Sub domain / Tenant work load
• Users to tenants mapping
• Orchestrate
• Tier - Throttle - Pattern

• Control Plane - Horizontal Services, Manage / scale. Services manage tenants
• Application Plane - Nature of SaaS Application, Isolation, Data Partitioning, Micro-Services

• Serverless - Control Plane
• Containers - App Plane
• Silo - Dedicated to Tenant
• Pool Side - Everything shared

• Silo - Deployment (Data Privacy / API Deployment)
• Serveless for Control Plane (Monitor / Measure)
• Onboard Architecture
• Self Service / Sign up
• Registration Service
• Create tenant / Routing / Policies
• Roles / Tiers / Isolation Policies
• Bill / Provision / Policies / Routing

• Provisioning experience
• Register - Provision / Self Service 
• Pooled Tenants

• Tenant Aware identity
• Access / Tenant Context
• Tokens with client context

• Tenant aware
• sub-domain
• which user pool to authenticate against
• Tenant management middle of execution

• Multiple / Global user pool
• MFA exclusive to user pool

• User pool mapped to

• User Management
• Admin - Tenants - Users

• Amazon Cognito - Implement secure, frictionless customer identity and access management that scales
• Billing providers - Third party
• Setup account with billing provider configure plans
• Send Activity data

• Metrics

• Application Plane
• Full Stack Silo/FUll Stack Pool

• Provisioning / Tiered 

• Full Stack Silo patterns

• Full Stack Pooled patterns

• Mixed models

• Multi-region

• EKS

• Isolation Implementation
• Full Stack Silo = Isolation off
• Full Stack Isolation
• Resource level isolation

 

• Deployment driven isolation

• Runtime enforced isolation

Keep Exploring!!!

Learning Data Science related topics

Many times I see a lot of Learning Data Science related topics

Some of the things I still learn and keep learning for work, and personal learning are at all levels: advanced - recent trends.

I tried to learn all again and again all below topics :)

• Maths fundamentals
• Differential calculus 
• Basics Linear algebra, matrix decomposition, pca
• Algo fundamentals, Trees, entropy, svm kernels, higher dimensions representations
• Feature engineering, boosting, bagging
• backprop basics, cnn, pooling, convolution, activation, dense, softmax
• CNN network design, loss functions, gradient descent techniques, Transfer Learning
• RNN, LSTM
• Transformers, encoders, decoders
• Basics of NLP, NER, Preprocessing, CRF, Naive Bayes, Sentiment Classification, NER Custom detection, Topic Mining
• Vision - detection, classification,segmentation
• Forecasting - Time series, linear regression, TFT, GluonTS
• Recommendations - Basics, Apriori, User-User, Item-Item, Recency-Frequency-Value, Ranking, Contextualize,Realtime, Batched
• MLOps, Deployment, FastAPI, AWS Lambda, SaaS Approaches, MLFLow, Dockerize, Streamlit, API

This is on top of other DB stuff :).

• Learn to connect on top of what you already know. 
• Grades, certifications does not mean expertise
• To grow, compile code, build your solution, copy and run existing code all are different levels of skills and expertise...

Learn at our own pace...Ever better...

It takes time to know 'Why it works, How it works, How it works when I debug Step by Step'

Be authentic to yourself, that's all Life is :)

• Anxiety is built into the very nature of leadership. It can—and should—be harnessed into a force for good. 
• Disrupting our anxiety starts by recognizing that anxiety is just data and it exists on a spectrum

Ref - Link 

Keep Going!!!

Personal assistant + Health App + Recommendations

 

Ref - Link

Keep Exploring!!!

Stability_AI Image generation

Prompt

• Beach blue checked dress for frmale adult

Code

Results

Keep Exploring!!

Estimates vs Projects vs Big picture

Sometimes I also have a mixed bag of tasks, Estimates depends on many factors, Predictable estimate vs just estimates 

• Vision papers / Experiment colab codes
• API creation
• Dockerize
• AWS / Azure learn sassify
• DB Tasks
• Catching up with GenAI 
• Mapping products vs models vs MVP
• Sell what you already have built
• POV / Proposals / PPT / Architecture

If I were a freelancer I would focus on building on some products and be a seller / reseller. Being an Expert of all takes time. Coding vs Full Stack vs ML vs DL vs Ideas .. Learning whatever is possible but need to pick and choose

Keep Exploring!!!

 

Automated image quality with cleanvision

What aspects are automated?

dark, light, odd_aspect_ratio, low_information, exact_duplicates, near_duplicates, blurry, grayscale images 

Keep Exploring!!!

Sample python code vs Kubeflow pipeline.

 

When a pipeline is submitted, Kubeflow creates a Kubernetes pod for each step in the pipeline. The pod is responsible for running the code associated with the step. The code is typically packaged as a Docker container, which is then deployed to the Kubernetes cluster. The pod is responsible for downloading the container, running it, and then reporting the results back to the Kubeflow Pipelines platform.

The Kubeflow Pipelines platform is responsible for managing the execution of the pipeline. It will monitor the status of each step and ensure that the steps are executed in the correct order. It will also handle retries and rollbacks in the event of a failure.

DSL.Pipeline is a domain-specific language (DSL) for creating and managing Kubeflow Pipelines. It provides a way to define a pipeline as a set of steps, each of which is a container image. It also provides a way to define parameters and artifacts that are shared across steps. DSL.Pipeline is designed to be easy to use and to provide a consistent way to define and manage pipelines.

DSL.ContainerOp is a Kubeflow Pipeline component that allows users to run a containerized workload. It is a wrapper around the Kubernetes Pod API and provides an easy way to define and execute containerized workloads. It allows users to specify the container image, environment variables, command line arguments, and other parameters that are needed to run a containerized workload. It also provides an easy way to define and execute containerized workflows.

The above pipeline code creates a Docker image from the Python script. The pipeline code defines a function called add_two_numbers_pipeline, which takes two parameters a and b. It then creates a ContainerOp, which is a type of operation in Kubeflow Pipelines. This ContainerOp defines the Docker image to be used, which in this case is python:3.7. It also defines the command to be run, which is a Python script that calls the add_two_numbers function with the two parameters a and b. Finally, it compiles the pipeline into a YAML file.

The pipeline code does not actually generate a Docker image, but instead creates a definition of the Docker image that can be used to generate the actual Docker image. Kubeflow Pipelines does not create Docker images in runtime. It uses existing Docker images to run the pipeline steps.

Keep Exploring!!!

Pulumi vs. Terraform

Pulumi vs. Terraform

Pulumi’s universal infrastructure as code platform helps teams tame the cloud’s complexity using the world’s most popular programming languages (TypeScript, Go, .NET, Python, and Java) and markup languages (YAML, CUE).

Terraform provides open-source infrastructure as code software for cloud service management with a consistent CLI workflow. Terraform allows you to write, plan, and apply changes to deliver infrastructure as code.

Keep Exploring!!!

ChatGPT - LLM - Vector databases - Document Search - Engines

Text is getting revolutionized everywhere. Search, Summarize, Content enrichment, etc..

• Azure Cognitive Search, you can search through millions of documents or data points.
• Azure OpenAI service uses its language modeling capabilities to understand the question and the context provided by the search engine, to generate an answer in natural language.

Ref - Link

Keep Exploring!!!

Vision Startups - Headcount

Ref- Link

Keep Exploring!!!

Ref - Link

Session Notes - Building the Future with LLMs, LangChain, & Pinecone

• Train models on datasets
• Transformers models and better results

• LLM out of box will do it

• LLM one-stop solution for NLP
• Summarization, Chatbots, Question, and Answering
• Key components

• Indexing, Search, LLM, Prompts all of them put together
• LLM models take actions 

• Pair embeddings in vector database helps in semantic search

Ref - Link

Keep Exploring!!!