AI & ML Workshops
Discourse surrounding artificial intelligence is full of interference. Companies who can distinguish the noise from information are able to navigate through the ambiguity created by the prevalence of AI in the marketplace. Our engineers have established a workshop series around a proven set of methods for successfully implementing production-grade AI and ML workloads. We believe AI can be approachable, explainable, and easy to use. The foundational workshop is built with conventional software development teams in mind.
Workshop Outcomes
This workshop focuses on three goals: centralize AI leadership, create an AI adoption road-map, and enable technical resources with tools and patterns. At the conclusion of the workshop, the following items are delivered.
- A centralized AI Center of Excellence.
- A formalized AI adoption epic road-map.
- A live deployment of the workshop reference architecture.
- Interactive and non-interactive course material.
- Continued support for AI adoption and expansion.
Adoption Advocacy
Inspired by Amazon’s Cloud Center of Excellence, workshops begin withthe formation of an AI Center of Excellence. This is a group or team that leads the organization in AI adoption. The AICoE provides guidance on best practices and governance policies. This team is created to give stakeholders responsibility, and provides an opportunity to identify champions and advocates.
Interactive Environment
Tools for demonstrating ML training and development capabilities within AWS can be deployed to provide an interactive environment for workshop participants. This environment implements AWS’s AI Roadmap and provides a suitable foundation for the creation and operation of production-grade models.
Interactive Workshop Material
Workshop materials are provided as a set of interactive Jupyter Lab notebooks. This allows participants to follow along and experiment in a live environment while receiving guidance from instructors. Zybe offers AWS, Azure, GCP, and self-hosted (Docker, Kubernetes) variants of these notebooks.
Course materials are also provided as LaTeX, a high-quality document preparation system used for the communication and publication of scientific documents. This material is easily re-branded, or converted to other formats (such as Markdown) suitable for embedding in a corporate knowledge-base.
Workshop Schedule
Workshops are broken into business and technical segments, focusing on strategy and outcomes with leadership, and concrete tools and patterns for developers. Below is an example of our simplest workshop for establishing fundamentals.
Technical Workshop Modules
| Module | Description |
|---|---|
| Pre-Processing | Practical tools and patterns to gather and prepare data for model training and validation, with a focus on cataloging and quality assurance. |
| Training | A foundation for automating the training and evaluation of models, and an exploration of foundational models. |
| Evaluation | Evaluation and explainability tools allow developers to gain a mechanistic understanding of model internals. |
| Management | Demonstrates how to integrate models that have passed evaluation into production environments while ensuring their reliable operation. |
Readiness Assessment
Before beginning a workshop, Zybe instructors work with stakeholders to assess where they are in their AI/ML adoption journey. This assessment is used to gauge the level of complexity surfaced to participants in workshop material.
- Teams seeking to establish fundamentals.
- Teams familiar with using foundational models.
- Teams learning to develop custom models.
- Teams seeking to scale ML workloads for production.
Language Ecosystem
Machine Learning can be implemented in many languages, however, we recommend that teams seeking to operate effectively with industry standard tools adopt some mixture of Python and Java.
Team Composition
We recommend a minimum complement of five dedicated resources for these workshops. In many cases additional MLOps engineers are needed to support this core team. This team is often assisted by IT and other internal departments as needed.
