Data Science UA Meetup: From prompts to LLMs and synthetic audiences – How AI works today
Join the offline meetup by Data Science UA – From Prompts to LLMs and Synthetic Audiences: How AI Works Today.
The event will be valuable for AI engineers, data scientists, product managers, developers, and anyone eager to understand how modern AI approaches really work in practice. The session will be moderated by Alexandra Boguslavska, CEO and Founder of Data Science UA.
Oleksii Shaldenko and Kostiantyn Zdor (Co-founders of Wantent) will lead a discussion on “The Future of Content Analytics: AI Agents, Synthetic Audiences and Video Attributes for Deeper Audience Insights”. Topics for discussion will include:
- What AI-driven approaches to video content analysis are there?
- How can AI agents simulate audience behavior and reactions, delivering fast, scalable insights?
- How to create and utilize synthetic audiences?
- What does in-depth analysis of content attributes (such as pace, color palette, emotional tone of scenes, and context) involve, and how does it enable more accurate predictions of engagement and content perception?
Maksym Kmet, Senior AI Engineer at MacPaw, will speak on the topic of “Prompts Are Just the Beginning: A Practical Guide to Controlling and Evaluating Language Models” and will share practical insights on:
- How language models generate the next token;
- How decoding strategies (top-k, top-p, temperature) shape the output;
- How to design effective prompts;
- How to evaluate responses using classical metrics and the “LLM-as-a-judge” approach;
- How to improve generation quality through automated prompt optimization.