Workers
The worker table contains one canonical record for each worker observed in the
processed employment events.
flowchart LR
raw[(US employment-position records)] --> select[Select one record<br/>per worker]
select --> workers[(Workers)]
classDef source fill:#e8eef7,stroke:#52739e
classDef process fill:#fff0cc,stroke:#a66b00
classDef data fill:#e2f2e7,stroke:#3b7a4b
class raw source
class select process
class workers data
Inputs
Build
| Component |
Location |
| Builder |
shared/data/build_processed.py |
| Runtime |
hct-tower-model-cuda-… (CPU execution) |
| Entry point |
python -m shared.data.build_processed --target employees / build-processed |
Output
| Dataset |
File / location |
Grain |
Key and variables |
| Processed workers |
employees_processed.parquet in /data/<coverage>/processed/ |
Worker |
Key: user_id Variables: highest_degree, sex_predicted, ethnicity_predicted, role_k150, user_location |
Command reference
Configured fields currently include highest_degree, sex_predicted,
ethnicity_predicted, role_k150, and user_location. For each field, the
builder keeps the first available value within the worker's ordered source
records.
The worker table supplies static employee features to model preparation and
worker-level characteristics to descriptive analyses. It does not define model
splits; worker splits are assigned later from the model configuration.