FastAPI Resume Builder MVP Development

I’m creating a proof-of-concept résumé generator aimed at veterans transitioning to civilian careers and I need a Python specialist to turn the concept into a clean, test-driven MVP.

Web app (Streamlit)
• Single entry-point module, app.py, that renders forms for contact details, MOS codes, and target roles.
• Real-time translation of each entered MOS into civilian skills/keywords, with a live preview the user can export as JSON.
• Integrated AI: auto-generate résumé summary (2–3 sentences) and STAR-style bullet options with “Regenerate/Apply” controls (editable in UI).
• Strong Pydantic v2 validation on every submit; no auth or persistence at this stage.
• DOCX download from the UI (ATS-friendly template). (PDF not required for MVP; add only if trivial within scope.)

Command-line tool
• build_resume.py reads the JSON profile and, via Jinja2 + docxtpl, outputs a polished DOCX using the same AI + mapping logic as the UI.
• Templating is easy to extend and supports simple branding tweaks (logo, color palette, accent color).

Architecture & quality gates
• Clean separation of concerns: core/ modules for mapping_service, resume_service, and ai_service (provider-agnostic; env-flagged; mock provider for tests).
• Deterministic output: stable ordering, fixed template styles, and a golden-text test (extract DOCX → normalize whitespace → compare).
• Thorough tests with pytest + coverage; lint/type checks with ruff/black/mypy.
• Inline docstrings and a concise README (setup, run commands, how to add MOS mappings/templates, AI config).
• Privacy: no PII persisted; all data in session/memory only.

Deliverables
1. app.py Streamlit project with minimal CSS/theme assets
2. build_resume.py CLI and one ATS-friendly docxtpl template (templates/classic/Resume.docx)
3. Test suite (unit + integration with mock AI) and coverage report
4. README and sample inputs (data/mos_mapping.csv + profile.sample.json)

If this aligns with your expertise in Streamlit, Pydantic v2, Jinja2/docxtpl, and DOCX generation/testing, please share a brief architecture/testing approach (including DOCX determinism + AI mock), examples of prior DOCX/Jinja work, and a fixed price that fits a ~20-hour MVP.