From Notebook to Production: Building End-to-End AI Systems

In this episode, we talk with Mariano Semelman about what it takes to build end-to-end AI solutions in production, especially in messy, high-scale e-commerce environments where a good model alone is not enough.
Mariano is a Lead Data Scientist and Machine Learning Engineer with more than 10 years of experience in e-commerce. At OLX, he works on media solutions for sellers, including systems that turn a user’s product video into a structured marketplace ad by automatically selecting images and extracting relevant details.

We’ll discuss:

• Why data scientists should learn to operate their own models
• How tools like Claude Code and Cursor can help bridge the gap between research and production
• What it takes to build video-to-ad systems that are reliable and scalable
• How to think about extensibility, monitoring, and maintainability in AI products
• When to prioritize customer impact first and pay down tech debt later

About the Speaker:
Mariano Semelman is a Lead Data Scientist and Machine Learning Engineer with over 10 years of experience in e-commerce. He focuses on building end-to-end solutions and explores how agentic tools can make the path from research to production more seamless. At OLX, he works on AI-powered media solutions that simplify the listing process for sellers.

**Join our Slack: https://datatalks.club/slack.html**