// Sovereign Machine Intelligence Directorate

ALPHA Teaching Portal Administration

Manage Amharic language teachers, review active session hours, authorize payment records, and approve verified training datasets.

shield

Administrator

Level 5 (SecOps)

Initial Teaching Budget

600.00 ETB

Allocated instruction pool

Approved Teaching Expenses

0.00 ETB

0.00 hours authorized @ 25 ETB/hr

Remaining Budget Balance

600.00 ETB

Ceiling enforced server-side

Pending Review Sessions

Est. 0.00 ETB awaiting approval

Pending Teacher Applications

Applications remain pending until explicitly approved by an authorized administrator. Approving provisions a teacher account with secure credentials.

Applicant & Phone	Introduction & Linguistic Background	Consent	Submitted	Status	Actions
Loading applications...

Registered Language Instructors

Active teachers authorized to communicate with ALPHA at 25 ETB / approved teaching hour.

Teacher	Phone / Login	Sessions	Approved Hours	Approved Earnings	Status	Access Control
Loading teachers...

Teaching Sessions & Payment Authorizations

Review active teaching durations recorded by the server-side anti-tamper heartbeat. Approving calculates payment at 25 ETB/hour against the 600 ETB budget.

Session Code	Teacher	Duration (Active)	Messages / Corrections	Calculated Payment	Status	Review & Action
Loading teaching sessions...

Learning Data Verification

Only approved, quality-checked corrections enter ALPHA's PyTorch training corpus. Raw Amharic text is strictly preserved.

Prompt & Instructor	Prior ALPHA Output	Teacher Correction	Explanation & Focus	Status	Action
Loading learning data...

PyTorch Training Dataset Export

Export verified human teaching pairs directly into the existing ALPHA PyTorch training pipeline (`c:\Users\user\Documents\alpha\train.py`).

description

Text Ingestion Format (.txt)

Exports raw Amharic conversation pairs formatted for direct ingestion into c:\Users\user\Documents\alpha\data\processed\train.txt.

download Download alpha_amharic_train.txt

data_object

Structured Metadata (.json)

Exports complete dataset with session IDs, teacher IDs, linguistic rules, explanations, and verification timestamps.

download Download alpha_amharic_dataset.json

Note on Model Training: Exporting the dataset does not automatically alter ALPHA's weights. Model training is executed locally on the PyTorch pipeline using python train.py to ensure deterministic evaluation and loss calibration.