The machine learning models are valuable only when deployed, and we can take full advantage of our business use case. When we start machine learning model development, we mostly focus on which algorithms to use, feature engineering, and hyperparameters to make the model more accurate, but the model deployment is a most critical step in the machine learning pipeline.

In this workshop, we are going to learn about ML lifecycle from gathering data to the deployment of models. Researchers and Data Scientists can build a pipeline to log and deploy machine learning models. We will learn about machine learning models’ challenges in production and different toolkits to track and monitor these models once deployed.

FORMAT: Pre-recorded videos (More than 5.5 hours of content) & Colab notebooks

Pricing: $19.99

Note: the presentation used in the workshop will not be shared with participants.

Get the workshop videos at $19.99

 

After payment, you will receive an email with a download link to the whole workshop (Videos & notebook links).

ADaSci Members receive a 30% discount.

The workshop is free for CDS Charterholders.

Content

Challenges Deploying Machine Learning Models

  • Traditional software development vs machine learning
  • Large data size
  • High-quality data
  • Geography dependencies of data
  • Computation power
  • Data privacy protection
  • Bias detection and mitigation
  • Result transparency and explainability
  • Trust 

Machine learning lifecycle

Data management

  • Data collection 
  • Data preprocessing 
  • Data augmentation
  • Data analysis

Model learning 

  • Model selection
  • Training
  • Hyper-parameter selection

Model verification 

  • Performance metrics
  • Formal verification for Regulatory

Model deployment

  • Integration
  • Monitoring
  • Updating

Ethics and Security of the model 

Tools for Machine learning Tracking 

  • Hands-on MLflow.org
  • Hands-on Neptune.io
  • Hands-on Comel.ml
  • Hands-on wandb.ai 

Model Compression before deployment 

  • Tools for model quantization
  • Hands-on TensorFlowLite
  • Tools for model pruning 
  • Hands-on model pruning 

Introduction to Model Deployment in Server and Serverless Frameworks

  • Server vs Serverless Deployment
  • Introduction to Flask 
  • Receiving Data through GET and POST Request
  • Hands-on model Deployment in Flask
  • Introduction to Streamlit
  • Hands-on Streamlit
  • Introduction to AWS Lamda
  • Hands-On AWS Lambda

Model deployment on Edge device

  • Android 
  • iOS 
  • IoT devices
  • Serving model via REST APIs

Introduction to Continuous Integration and Continuous Deployment 

  • Whats is CI/CD
  • Setting up CI/CD
  • Setup CI/CD Config
  • Publishing the Model 
  • Testing the CI/CDPipeline

Post-production Monitoring 

  • Model Drift 
  • Error logs and Analytics
  • Retraining Pipeline due to environmental change

Tools And Techniques That Will Be Used

Flask, Streamlit, Mlflow.org, Neptune.io, Wandb.ai, Circle CI/CD.

Key Takeaways:

  • Hands-on experience on some of the most used tools for model tracking and deployment 
  • Certificate on Hands-on Deep Learning Model Deployment and Management
  • Learn about Flask and Streamlit 
  • AWS for AI/ML model deployment
  • Deploying models using batch, streaming, and real-time
  • Solutions to model drift and retraining pipeline for the deployed model

Prerequisites

  • Basic to moderate level python
  • Basic of Pandas, Numpy, Scikit-learn, machine learning and artificial intelligence  
  • Familiarity with Google Colab and GPU environment
  • Basic familiarity with object storage, databases, and networking

Required Tools

  • Any editor to run the python programs (preferably Google Colab Notebooks)
  • If working on an editor Pandas, Numpy, scikit-learn, TensorFlow, Pytorch and Keras must be installed.
  • High-speed internet connection

Your Instructor

Krishna Rastogi is Associate Director at Association of Data Scientist. He has experience in research & development, cutting edge engineering to develop products from idea to deployment. He comes with expertise in building computer vision applications using both hardware and software solutions in several domains.
He specialised in edge AI  domains and deploying deep learning models on small hardware devices without taking the raw data from the devices. He conceptualized, researched, and built 35+  product prototypes in Healthcare and Medtech. He presented some of his projects to Late Shri Dr. A P J Abdul Kalam and Shri Ratan Tata. He worked under Prof. Ramesh Raskar, MIT Media Lab, Boston as a visiting student for a year. He also worked with the MIT team to set up an innovation lab in Mumbai.

1 Comment
  1. Karan Mahesh Mankar 1 month ago

    Hi I am a graduate student at NYU and since I am planning on taking this workshop from the US, the timing is a little bit off for me. Will this session be recorded and emailed to participants?

Leave a reply

Your email address will not be published. Required fields are marked *

*

©2021 Association of Data Scientists

Log in with your credentials

Forgot your details?