Bespoke Datasets for Multimodal AI Products

Overview

This course is for product managers who want to build AI products that include multimodal data such as text, video, and audio. Leveraging existing data for AI applications is a great way to get started, but bespoke datasets can offer improved accuracy, domain specificity, and, ultimately, a competitive advantage.

The course starts with an overview of supervised machine learning techniques, computer vision—the technology that allows computers to "see"—and how computer vision and supervised ML can be combined. Then, you'll learn more about how to create bespoke datasets for computer vision and other applications, and conclude with a walkthrough of the key stakeholders involved in AI product development.

In the project, you'll develop a roadmap for an AI-powered video creation app.

Syllabus

• Supervised Machine Learning and Computer Vision
Dive deeper into each step of the supervised machine learning process, from data understanding to model evaluation, then learn how to combine these approaches with computer vision technology
• Data Annotation for AI Products
Learn how data can affect the performance of an AI system and determine whether a bespoke dataset is needed, then practice developing a data annotation plan
• AI Stakeholders and Roadmaps
Collaborate with stakeholders, enhance communication, and develop roadmaps for AI products. Develop skills in influencing without authority and the RICE prioritization framework.
• Roadmap for Tasty Cuts: An AI-Powered Video Creation App for Home Chefs
Build a roadmap for an AI product that combines computer vision and specialized data annotation needs

---

Taught by

Justin Leung, Kirubel Tesfaye and Brad Nemer

---

Subjects

Computer Science