LLMs generally reply in a nondeterministic format; it does not always comply with the formatting instructions given. This is where controlled generation (structured output) comes into play, where you ask an LLM to reply to comply with a given schema. In this post, you will learn how to use Gemini over Vertex AI and controlled generation to get structured output that follows a schema on job listings to summarize and categorize them. Let's get going!

I started writing the year-in-review (recap) posts from 2019, so this will be the sixth consecutive year I will write a yearly recap. Taking some time to reflect on things accomplished this year from a professional point of view, I think it will be a good rear mirror view to look back at 2024; let’s dive in!

Updating existing data is a core requirement of any web application; doing it efficiently will make your life easier. PostgreSQL, a robust and feature-rich relational database, offers a powerful and elegant solution for managing these updates: INSERT ON CONFLICT UPDATE. It is helpful to combine insert and update to Upsert and use the same logic for both operations. In this post, you will learn how to use INSERT ON CONFLICT UPDATE in Postgres to Upsert data effectively with practical examples. Let’s get going!