Abstract

As a Product Manager for AI tools, I see a lot of challenges arising when the uncertainties of AI world meet the well-defined processes of classical engineering projects. These challenges include:

• The nondeterministic nature of natural language inputs and outputs, that creates a lot of uncertainty and widens the gap between the working Proof-of-concept and a well-rounded enterprise level solution.
• Roles shift: the prompt engineering and testing effort highly relies on team members without engineering background (content writers, marketing professionals) who are experts in language but may lack familiarity with R &D processes.
• Speed vs. Quality: this old as time conflict creates more tension than ever when a company needs to adopt AI fast without compromising the high quality and security standards

All these things create a lot of friction in big projects, where previously there was a smooth and clear way of planning, execution, testing and delivery. In this talk, I'll suggest a few ways to optimize the work, to build good prompt evaluation processes, to teach project teams about prompt engineering and to build a process in which prompt engineering is a normal, well defined and somewhat predictable part of the workflow

Bio

Anastasia Podrabinek is a Product Manager at Wix, focused on data science and AI infrastructure. She brings years of experience building and managing scalable digital products.

Video