docs/team_roles.md

Roles, branches, and tasks (entire project)

Team

• Gemima Ondele
• Kachallah Fatima
• Mike-Brady
• Joseph Fabrice TSAPFACK

Branch assignment (primary)

• Gemima: feat/data-pipeline (data lead)
• Fatima: feat/evaluation (evaluation lead)
• Mike: feat/training (modeling lead)
• Joseph: feat/demo (demo and packaging lead)

Roles and tasks by week

W1 (Dataset pipeline)

• Gemima (data lead)
  • Implement download_data.py cleaning + filters
  • Export JSONL splits
  • Validate schema consistency
• Fatima (data QA + docs)
  • Define filtering thresholds
  • Produce dataset stats and short summary
  • Update README pipeline section
• Mike (data QA)
  • Verify split ratios and random seed
  • Smoke test JSONL loading
• Joseph (repo hygiene)
  • Maintain folder structure
  • Update CONTRIBUTING.md and doc links

W2 (Baseline + rubric)

• Gemima
  • Baseline generation script setup
  • Save sample outputs for review
• Fatima (evaluation lead)
  • Define test prompt set (50+)
  • Create human rubric (coherence, creativity, faithfulness)
• Mike
  • Configure baseline model settings (distilgpt2)
  • Document failure modes
• Joseph
  • Organize evaluation artifacts in repo

W3 (Fine-tuning)

• Gemima
  • Data loader and training-ready format check
• Mike (modeling lead)
  • Fine-tuning script (HF Trainer)
  • Track training curves and checkpoints
• Fatima
  • Prepare evaluation notebook/script for W4
• Joseph
  • Ensure reproducible run steps in README

W4 (Control + evaluation)

• Gemima
  • Add control fields (level/setting/tone) in prompts if used
• Fatima (evaluation lead)
  • Run baseline vs tuned evaluation
  • Summarize metrics and rubric scores
• Mike
  • Assist with controllability tests
• Joseph
  • Compile examples for report/presentation

W5 (Demo + report)

• Gemima
  • Final polish on data/scripts
• Fatima
  • Lead report writing (4-8 pages)
• Mike
  • Final model selection and results tables
• Joseph (demo lead)
  • Build Streamlit/Gradio demo
  • Prepare presentation slides