D8.7 Summary (3)
See the working text for this QS in
- #619_.docx.
- Demo URL there are actually many … this was a chaotic demo.
TOC
7.1 Tools Used — Full Process
Data Setup
Dataset — raw housing data uploaded/referenced in Foundry
Code Repository — the IDE environment where all code lives
Python Transforms (transforms.api) — used to clean, split, and prepare the dataset as a pipeline 1
Model Training
Code Repository (ML training template) — structured environment for ML work
scikit-learn — built and fit the regression model
palantir_models — defined the ModelAdapter and published the trained model as a versioned Foundry artifact 1
dill — serialized the model for storage and reloading
Inference Setup
Model Deployments — deployed the trained model behind a REST API for live inference 1
Functions (TypeScript) — wrote a wrapper function that calls the model deployment and returns a prediction 1
Functions Repository — the environment where the TypeScript Function was authored and published
UI
Workshop — built the front-end app with input widgets and a Metric Card 1
Workshop Variables — Static Numeric inputs wired to a Function-backed output variable
Numeric Input widgets + Metric Card — the UI components the user interacts with
7.2 Transform pipeline was only used for training
In your demo, yes — the Specifically it handled:
Feature engineering —
cleaning and preparing the housing data
Train/test splitting —
splitting the dataset before model fitting
Model training —
fitting the sklearn model and publishing it via palantir_models
The live inference (what happens when you change a slider in Workshop) bypasses the Transform pipeline entirely. It goes directly through the Function → Model Deployment path in real time.
That said, Transforms can also be used for batch inference — running predictions across a whole dataset and writing the results back to Foundry. 1 That just was not needed in your demo since you built a live, interactive UI instead.
So the short version for your tutorial:
Transforms =
data prep + training (batch, scheduled)
Function + Model Deployment =
live inference in the UI (real-time, on demand)
26.0903 (v1 26.0903)