Momenta Analytics
Your semantic layer, built from your SQL history
Start new thread

Momenta Analytics - Your semantic layer, built from your SQL history

by•
We automate semantic layer population from SQL histories. Why? Your analytics team knows how to query your data so they should teach your AI. Point it at your team's query history and it populates your semantic layer automatically: metrics, business rules, table joins, top query patterns. No manual tagging. Runs as a local Docker, so nothing leaves your environment. Exports portable, OSI-compatible files you own that feed Databricks Genie, Snowflake Cortex, or your own LLM. Free for 60 days.

Add a comment

Replies

Best
Maker
📌
Hey Product Hunt 👋 I'm Simon, co-founder of Momenta. Here's our origin story. My co-founder and I led analytics teams, and we hit the problem every analytics team does. At some point you get bottlenecked by your experts. The people who know the domain cold, translate it into SQL, and make the calls everyone else trusts ("we don't have a flag for a line-of-therapy switch, but here's a defensible way to do it"). They've been there a decade, nothing is written down, and you have nightmares about them quitting. Our first crack at it was nothing like Neuron. We thought the answer was auto-documentation, a self-generating Confluence. Wrong. What we couldn't let go of was the simple idea: the SQL our teams were writing was a repository of that specialized knowledge, and it is possible to recreate it from the SQL itself. Do that, and more people can run analyses, and your experts stop being a help desk for their own domain. We stayed convinced that was THE problem. We lived through Snowflake, Databricks, Tableau, all of it. Our companies spent millions on data. The bottleneck was never cluster size or missing data. It was always needing someone the organization trusts to sign off on an analytical decision, written as SQL. We think we've found the fit. The semantic layer is the emerging home for that knowledge. Fill it, and AI can take it from there. So we do the part we think is really hard and have gotten good at: turning executed SQL queries into semantic layer content. If it all goes right: you run Neuron on your query history, drop the output into your semantic layer, and your text-to-SQL AI starts performing like your best analysts. On all of our runs, that end-to-end process has never taken more than an afternoon. That's the motivation. It's day one, so I'd love your read, especially whether it holds up on your messiest, multi-step queries. I'll be in the comments all day. Free for 60 days, your output is yours to keep, your data stays on your machine.