About

Staff AI Engineer at Snowflake. I work on Snowflake’s internal Sales agent, a Snowflake CoWork agent that handles 12,000+ questions a day from 1,500+ daily users. I lead the agent’s CI/CD and evaluations, and I built the feature that lets users schedule recurring jobs and get alerts when thresholds are met. I collaborate with Snowflake’s product team, sharing CoWork feedback with them.

In my personal time I build my own developer tooling, which I open source when it’s ready. The first public release is nah, a guard that blocks catastrophic coding-agent actions before they run.

How I got here

Technical Support → Data Science → Applied AI.

I started as a Support Engineer at Cloudera, then moved to Snowflake. Solving the same tickets over and over got old, and getting fixes prioritized by engineering meant proving how often the same issues came back. That turned out to be an unstructured data classification problem.

So I moved to Snowflake’s Support Analytics team, where I built data products on years of historical support cases using BERTopic and embeddings. This was pre-ChatGPT, but as LLMs arrived my products improved: AI could summarize, classify, and extract insights from cases. I spent a few years presenting those findings to engineering and support leadership, helping decide what got fixed and what got built.

Now I build AI full-time.

This site

This site collects my public work on agents and developer tooling.

You can find me on GitHub, LinkedIn and X, or email me at manuel@schipper.ai.