SG Job Data — Summary + Projection App

• compute_summary.py — processes the CSV in chunks and writes summary.json with dataset summary and monthly salary series.
• app.py — Gradio app that shows dataset overview and runs a 5-year linear projection of average salary.
• requirements.txt — Python dependencies.

1. Create a Python environment and install dependencies:

1. Run the compute summary step (this may take a few minutes):

1. Run the app locally:

• Create a new Space (Gradio runtime).
• Upload repository contents including SGJobData.csv, app.py, and requirements.txt.
• Set app.py as the app entrypoint (default for Gradio Spaces).

• The projection engine in app.py fits a linear trend to the monthly historical mean average salary and extrapolates forward. This is intentionally simple — see compute_summary.py for monthly series creation. For production use, consider more robust time-series models (ARIMA, Prophet, or ML-based models) and per-occupation projections.

Hugging Face Spaces — recommended files to upload

• app.py — Gradio entrypoint (required by the Space to run the app).
• requirements.txt — Python packages to install on the Space.
• README.md — usage and notes (this file).

• SGJobData.csv.gz — compressed dataset (preferred over the raw CSV). app.py and compute_summary.py will detect and use the .gz automatically.
• summary.json — precomputed summary to avoid a long startup processing step.
• compute_summary.py — script used to create summary.json (useful for reproducibility).
• projection_engine.py — forecasting code used by the app.
• occupation_mapping.py — helper to map title → occupation categories.
• projection_snapshot.png — static preview image for the repo (optional).

• The compressed SGJobData.csv.gz dramatically reduces upload size and startup processing time. The app is already coded to prefer the gz file if present.

Space example configuration (instructions)

1. Create a new Space on Hugging Face with the Gradio SDK.
2. Set the hardware to CPU (the app runs fine on CPU); choose GPU only if you add heavy ML models.
3. Upload the files listed above. The simplest deployment requires at least app.py and requirements.txt.
4. The Space will automatically run app.py (Gradio). If you need a custom start command, set it in the Space settings.

Notes & cautions

• If SGJobData.csv.gz is large (tens of MBs to GBs), check Space storage limits and upload time; consider external hosting (S3/GDrive) and modify app.py to download at first run instead.
• Uploading summary.json avoids recomputing the whole CSV on first start — helpful if the Space should start quickly.
• Do not upload sensitive or private data. If the CSV contains PII, remove/anonymize before uploading.

Ready to help

• produce a trimmed sample CSV (e.g., 1% sample) for public demo Spaces, or
• modify app.py to download the compressed CSV from an external URL at startup, or
• prepare a short .zip or upload-ready bundle for you to push to Hugging Face.