Do you know about MLops ,Yeah you got it correct NOT DevOps it’s MLops
AI systems
don't just take time to develop.
Let's break
down what the setup and ongoing costs of an AI solution looks like.
In the AI world. We typically divide model development into two phases,
Model
training and Deployment:
Model training involves researching the problem, gathering the
appropriate data, selecting the right model type, and training it.
This journey is typically challenging as it exposes some organizational
weaknesses, like which concrete business problems AI can solve, or messy
data.
These
problems can be solved, but are usually ignored in the initial assessment.
Model training usually requires a significant amount of R&D from a business
standpoint, and also a technological one. From a finance perspective, we can
classify this as a capital expense.
Rather than
training your own model, you can also purchase or license models. After a model
is trained, it needs to be Deployed, and integrated with existing
systems.
Deployment:
This might
also require significant innovation depending on your business problem.
Privacy, latency, throughput, and network connectivity would all affect the
deployment pattern that you choose in the product configurations you would
have. Once the models and their infrastructure are up and running,
In most
modern applications, you'll need to continuously monitor and update your
models, which is typically referred to as MLOps. Like software, models
aren't static, and need to be updated and retrained. So these are the key
considerations for model setup and ongoing expenses. In the next video, we'll
discuss selecting the right model for the task.
Author
Techsourcing ML Team
Comments
Post a Comment