mlops

This is the code repository associated with my development notes notebook.

I need it to improve development efficiency. I study concepts, prototype solutions, save them here along with their comments. When needed, I know where I can quickly find code examples to reuse.

Simulation of a real hospital scenario with a ML model in production

This repository showcases an end-to-end machine learning project implemented with a focus on MLOps.

MLflow deployment plugin for Knative targets

Fork of Rubrix for natural language processing machine learning dataset labeling and RLHF (reinforcement learning human feedback) and active learning. Original license was Apache 2.0 which allows Argilla.io to modify the software without sharing it with the community.

A common use case for AI and ML in security is to help establish a baseline of normal operations and then alert a team to potential anomalies. Also, AI and ML can also be used to improve operational effectiveness by identifying the more mundane tasks that people are doing all the time. The technology can create or suggest automation playbooks that will save time and resources.

Milliways menu for mlops

A generic implementation roadmap that can facilitate MLOps for any ML problem in detail. This roadmap is intended for reducing problem solutioning, and using problem solving instead.

Security is a never ending story. Tips and frameworks for potential solutions and raising discussions about topics like data access, privacy, human intervention etc.

Logging, monitoring, and metrics are so critical that any model getting shipped intoproduction must implement them at the risk of hard-to-recover catastrophic failure. High confidence in robust processes for reproducible results requires all of these components.

For ML operations, it is critical to isolate responsibilities as much as possible from the process of getting models into production. Isolating components can then pave the way for reusability elsewhere, not only tied to a particular process for a single model. An excellent solution for creating microservices is using serverless technologies.

This is a minimalist MLOps-style project to bring together the basics of cloud computing and how to set up and use CI/CD.

AWS is the largest cloud platform. There are many different ways to approach a problem using AWS technology. These are some well-travelled roads and some unique corners of AWS.

Azure is already solving challenging problems related to operationalising machine learning, from registering and versioning datasets to facilitating monitoring and deploying live inferencing models on scalable clusters.

It is becoming fairly common to see models getting deployed on mobile phones and other (small) devices that you can plug into any computer with a USB port. The problems that edge inferencing provides (like offline, remote, and fast access) can be transformational, specifically for remote regions without access to a reliable source of power and network.

The number one advantage of using the Google Cloud may be that its technology is ideal for a multicloud strategy: Exploring GCP for use in MLOps.

ONNX is lowering the friction that exists to deploy to remote environments where it is impossible to connect to the internet or have any network connectivity.

AutoML, the automation of modelling, is an essential capability for doing MLOps. AutoML improves the ability to push models into production, work on complex problems, and ultimately work on what matters.

Fast scaling of resources and transitioning deployment environments from cloud providers, or even moving workloads from on-premise (local) to the cloud is far easier to accomplish with containers.

this project it to practice all concepts and knowledge in the course mlops-zoomcamp