CanOfWorms/Entrepreneur-Readyness

Entrepreneur Readiness — Confidence Model (v0)

A lightweight regression model that scores how ready an individual is to be an entrepreneur from free‑text input. Score is in [0, 1] (we also show 0–100 in the app).

What this is

• Model type: Text regression (Transformers AutoModelForSequenceClassification with num_labels=1 + problem_type="regression").
• Output: Single float in [0,1], interpreted as readiness confidence.
• Baseline backbone: distilbert-base-uncased (you can swap in any encoder).
• Dataset format: JSONL with fields text (string) and label (float in [0,1] or [0,100]).

---

Quickstart

0) Environment

python -m venv .venv && source .venv/bin/activate  # on Windows: .venv\Scripts\activate
pip install -r requirements.txt

1) (Optional) Inspect the tiny sample dataset

data/entrepreneur_readiness.sample.jsonl

2) Login to Hugging Face

huggingface-cli login

Make sure you have a HF account and a write token with read+write scope.

3) Train

python train.py   --model_name distilbert-base-uncased   --train_file data/entrepreneur_readiness.sample.jsonl   --eval_file  data/entrepreneur_readiness.sample.jsonl   --output_dir ./outputs   --epochs 3   --batch_size 8   --lr 5e-5   --hub_model_id your-username/entrepreneur-readiness   --push_to_hub

Notes:

• Recommended repo name style on HF: all lowercase with hyphens.
• Start the repo as private. Flip to public when you’re ready.

4) Inference (local)

python inference.py --model_id your-username/entrepreneur-readiness --text "I have a validated problem..."

5) Gradio App (local)

python app/app.py --model_id your-username/entrepreneur-readiness

Open the printed http://127.0.0.1:7860 URL.

6) (Optional) Push a Space

Create a new Gradio Space named your-username/entrepreneur-readiness-app and upload:

• app/app.py (rename to app.py in the root of the Space)
• requirements.txt

Then set Space SDK to Gradio.

---

Data format (JSONL)

Each line is a JSON object with keys:

{"text": "free text", "label": 0.84}

Labels may be given as [0,100]; the trainer normalizes to [0,1] automatically.

Labeling guide (suggested)

Label in [0,100] using the holistic rubric below, then divide by 100:

• 90–100: Clear validated problem, engaged users, revenue or strong pre‑orders, runway ≥ 9 months, strong execution cadence.
• 70–89: Solid plan and early validation, some funding/runway (6–9 months), MVP near-ready or in pilot.
• 50–69: Plan forming, limited validation, learning mindset, early network/customer discovery.
• 30–49: Idea still fuzzy, little validation, unclear runway, limited execution plan.
• 0–29: Very early stage with major unknowns; high risk and minimal readiness signals.

Consider factors: customer discovery, problem validation, founder–market fit, execution discipline, financial runway, resilience, network.

---

Ethics & Safety

This model provides an opinionated readiness score and can be wrong. Don’t use it to gate access to opportunities. It should not replace human judgement; treat it as a coaching heuristic.

License

Code: MIT. You are responsible for your data licensing.