The data and ML stack we wished we had
Heimdall warehouses data in Lake, explores it in Lab, and ships ML, Forecast, Loop, and Resolve as production APIs — so you are not waiting on a data science queue to get a model into an app.
Our Mission
Help teams go from files and databases to a monitored prediction API without standing up a lakehouse program. We build Lake, Lab, ML, Forecast, Loop, and Resolve as one path — catalog, train, deploy, or answer text you define — so the data you already have can answer a question in production.
Guided by our Core Values
These principles shape everything we do and guide our decisions
Innovation
We would rather ship a working bronze-to-API path than wait for a perfect stack. That means trying new modeling approaches, and dropping the ones that do not earn their keep.
What it means
- Learn from production, not just papers
- Solve today's warehouse and modeling bottlenecks with tools teams will actually use
- Experiment in the open, and kill ideas that do not help customers ship
Accessibility
A lakehouse-plus-MLOps program should not be the price of a prediction API. Lake, Lab, ML, Forecast, Loop, and Resolve exist so smaller teams can catalog data, ship models, and answer text without a specialist queue.
What it means
- The same products should work for a first CSV and for a production database
- We build with customer feedback, not a one-size-fits-nobody platform tour
Simplicity
Machine learning is useful when someone can train, understand, and call a model. We design for that path — gold datasets, no-code training when you want it, Python in Lab when you need it.
What it means
- Interfaces should make the next step obvious
- Explainability and monitoring belong next to deploy, not in a separate project
Meet Our Founder
Joel Reji
Founder & CEO
Joel Reji founded Heimdall after years as a machine learning leader and data steward at Bank of America. He holds a Master's in Data Science and Business Analytics and is a named inventor on a machine learning patent. Heimdall exists to make warehouse-to-API modeling available to teams who should not need a platform program to start.
Ready to put a model in production?
Bring data, or start without it.
ML and Forecast train on gold from Lake and Lab. Loop and Resolve start with no dataset, including ticket routing and urgency scoring.