MLOPS Training – Running successful AI projects in Production

Kurssin esittely

Why MLOPS? 

MOVING AI WORKLOADS RELIABLY IN PRODUCTION
”87% of projects don’t get past the experiment phase and therefore, never make it into production..”

”Companies that are holding on from deploying algorithms and models in production, stand a lot to lose.

Think about it: If an AI model is to save a company $10M a month, then not maintaining it will mean the company loses $10M a month!

Unlike static resources such as physical machines and even some databases and software, machine learning models are a constantly moving target. They require continuous monitoring and continuous deployment. On top of this, as the clients ecosystem changes in the real-world, so does the machine learning model.

MLOps provides an opportunity for companies to actively guard and combat data-shift and subsequent data-decay of their model performance by constantly feeding new data and features to it. This active intervention allows for superior model performance, meaning greater data security and accuracy, and that allows businesses to develop and deploy models at a faster rate.” –

Tarry Singh, CEO deepkapha.ai

MODULE 1: AI IS SOFTWARE 2.0
Here you will learn about:

AI is increasingly driving change in enterprise. Here you will learn about the pre-AI infrastructure and pre-AI software development paradigms. Then you will learn about how consistent liquefaction of technology stack has happened. First it was infrastructure which led to the birth of cloud computing. This was the software 1.0 era. Now AI is driving the Software 2.0 era where low-code and no-code applications are becoming popular with citizen developers.

 

MODULE 2: MLOPS – AN INTRODUCTION
Here you will learn about:

MLOps is all the engineering pieces that come together and often help to deploy, run, and train AI models. With that, you will invariably encounter (atleast) the following tightly interwoven components of MLOps: DataOps – Data Engineering, Model Ops – Machine – & Deep Learning and DevOps – Software engineering required to develop elegant apps and solutions.

You will learn about the MLOps concepts, Benefits such as Benefits such as rapid innovation through robust machine learning lifecycle management, creating reproducible workflow and models, easy deployment of high precision models in any location and finally management and monitoring of the same for sustained value creation.

 

MODULE 3: AI SOFTWARE &  APP STACK
Here you will learn about:

  • Available tools in the market today – COTS (commercial off the shelf) as well as Open Source.
  • MLOPs Design Elements: You will understand about each dimension of the MLOPs domain
  • Featuritis*: The curse of more features
  • App Engines – What are they?
  • What is an App Engine Framework – Standard vs Flexible?
  • Public Cloud App Engines: Google App Engine – How it works
  • AWS App Framework – How it works?
  • Azure App Framework – How it works?
  • On-Premise App Engine: Build Your Own
  • A comparative presentation of on-premise vs BYO approach and how to benefit from best of both worlds approach.

 

MODULE 4: MODEL SERVING
Here you will learn about:

  • Model Serving: What is it and when do you need it?
  • Techniques Used: Stateful vs Stateless Serving
  • Batch Serving: How to do it?
  • Continuous Model Evaluation: How is it done? More?
  • Software Tools and Architectures: Apache Spark, Kafka, Beam, DLHub & more.
  • Summary: Familiarize with some use-cases from the industry that focus on lessons learned and the strategies used to overcome them.

 

MODULE 5: MODELS INFERENCING, INTERPRETATION  & EXPLAINABILITY
Here you will learn about:

  • Model Inferencing: What is it and when do you need it?
  • Online Inferencing: Designing for Cloud Based environments, Implementing Online Inferencing, Optimization – Do we have a fix for featuritis*?,
  • Releasing Value inside your Organization: CI/CD(Deployment, Rollout), CM (Continuous monitoring)
  • Batch Inferencing: Implementation, Pros and Cons of Batch Inferencing, Software Tools for Batch inferencing
  • Lessons Learnt – Some Do’s and Don’ts


MODULE 6: REGISTERING YOUR MODEL
Here you will learn about:

  • Model Registries – What is it?
  • Implementing your own Model Registry
  • Model Registry API and its benefits: East of use, repeatability and ease of testing.
  • Designing & Implementing your own
  • Model Registry API
  • Reliable CI/CD: Serving Reliably Trained Models in Production Continuously (CI/CD)
  • ModelOps: Model Train/ Model
  • Iterate / Model Retrain
  • Promoting to Production
  • Retrieving for Inferencing
  • Software Tools for Model Registries


MODULE 7: EMERGING AI MICROSERVICE ARCHITECTURES
Here you will learn about:

  • Advantages of deploying microservices architecture: scalability, speed, fault isolation
  • RESTful architecture for ML/AI
  • Advantages for using RESTful architectures
  • Use cases Explained
  • FARO – Learn with us. We will use our own stack – a modified FARM approach.
  • Other approaches and tools

 

MODULE 8: PERFORMANCE TESTING  STRATEGIES
Here you will learn about:

  • Model Testing – What is it?
  • Performance Testing (before Rollout)
  • Dataset Representation – A Moving Target
  • Combating Data Leakage
  • Fighting Hidden Feedback loops
  • Performance Issues in SubPopulations
  • Example Uses cases from Industry Verticals
  • Software Tools for Performance Testing
+ Lue koko kurssiesittely

Kouluttaja:

Koneoppiminen, Tensorflow

TARRY SINGH

Chairman & CEO of DK Ventures.

Tarry Singh is Chairman & CEO of DK Ventures. He is cofounder and AI Researcher of AI startups deepkapha.ai , curae.ai , a healthcare AI startup. deepkapha.ai is part of NVidia Inception Program for leading AI startups worldwide.

Tarry has over 20 years of experience working with data and has advised CxOs, Leading Government Authorities (including Ministry and President Level) of global organizations and country states to setup data-driven organizations from scratch.

He speaks regularly at global AI leadership summits worldwide and conducts workshops on a regular basis with his TAs who are currently PhDs in various disciplines such as NLP, Computer Vision, Robotics and other Artificial Intelligence disciplines.

Lue lisää

MLOPS Training – Running successful AI projects in Production

Teema:
Ketterä kehitys
Kouluttaja:
TARRY SINGH
Aloituspäivämäärät:
  • 11.11.2021
Kieli:
English
Kesto:
2 päivää
Hinta: 1900€+alv
Ilmoittaudu kurssille

Tuleeko yrityksestäsi useampi osallistuja?

Järjestämme myös yrityskohtaisia kursseja.

Haluan koulutuksen koko yritykselleni

Saattaisit olla kiinnostunut myös näistä

Datan hallinta

Practical Data Governance, Stewardship & Compliance

Lue lisää
Datan hallinta

Data Management Fundamentals and DAMA Certification Preparation

Lue lisää
Tiedolla johtaminen

Data Strategy – 4 Day Evening Course for Leaders

Lue lisää
+