Skip to content

Home

Jump to bottom
Kavya Sri Meka edited this page Jul 18, 2026 · 1 revision

Welcome to the CulRag wiki!

Perfect. You're in a clean state: main is current, tests passing, all three phases + paper scaffolding committed, ready to hand off.

Checkpoint Summary

Phase Status Commit
Phase 1 (RAG pipeline) ✅ Complete On main (PR #23 + merge)
Phase 2 (Evaluation) ✅ Complete On main
Phase 3A (Dashboard) ✅ Complete On main
Phase 3B (Paper scaffolding) ✅ Complete On main
CRC32 fix (determinism) ✅ Complete On main (merged)
Paper 1 Methods draft ✅ Complete On main
Paper 2 Protocol + Kit ✅ Complete (draft) On main
Tests ✅ 36/36 + 6/6 Passing

Send to Vartan

The task list is ready to paste. Frame it as:

Vartan — CulRAG Pilot Study Prep (Week of Jul 20)

Everything is on main (GitHub link). All tests passing. Here's what we need from you:


[Paste the task list above]


Timeline:


    

  • Data items 1–5: Due end of Week 3 (unblocks Paper 1 results)
    • 

  • Paper sections 6–8: Draft by end of Week 4
    • 

  • IRB prep items 9–11: Review by end of Week 2 (blocks recruitment Week 8)
    • 

  • Sunday sync (Item 12): Decisions on data merge, recipes, deployment
    • 



Let me know if you hit blockers on the IFCT 2017 licensing or can't find papers for the TODO verify entries.

Your Work This Week (Kavya)

Immediate (Days 1–2)

  1. Confirm IRB status

    • Email UNT Office of Research Compliance:

      "We're planning a 4-week pilot study with 8–10 human participants (self-reported weight, optional feedback, SUS survey). Does this require IRB approval? If yes, what's the timeline and what documentation do we need?"

    • CC Vartan (he may own this at Datara legally, or UNT handles it)

  2. Update GitHub Projects (Kanban board)

    • ✅ Phase 1: Done
    • ✅ Phase 2: Done
    • ✅ Phase 3A: Done
    • ✅ Phase 3B: Done
    • 🔄 Paper 1 Writing: In Progress (you own; target Week 6 submit)
    • 🔄 Paper 2 IRB: In Progress (blocks Week 8 recruitment)
    • 🔄 Data Integration: In Progress (Vartan owns; target Week 3)

Week 1–2: Paper 1 Writing

Expand stubs in paper/:

  1. Abstract (250 words)

    • Expand the outline → prose
    • Keep placeholder metrics: "X% precision@5", "Y% recall@10"
    • Cite 3–4 papers already in .bib

  2. Introduction (800 words)

    • Expand outline → prose
    • Add ICMR/WHO citations for India health burden (Vartan's task #6, but you can draft around placeholders)
    • Focus: Why cultural grounding matters + gap in existing systems

  3. Related Work (1,200 words)

    • Expand outline → prose
    • Cite papers from references.bib (Vartan will verify TODO verify entries)
    • Focus: RAG + nutrition + cultural AI landscape

Skip for now: Results, Discussion, Conclusion (wait for real data from Vartan's metrics run, Week 3–4)

Week 2: IRB Inquiry + Initial Protocol Review

  1. Send IRB email (Day 1)
  2. Review paper/protocol.md + recruitment_kit/ for placeholders you need to fill:
    • Your institution name
    • IRB contact details
    • Survey platform (Qualtrics? Google Forms? Paper?)
    • Data retention period (1 year? 3 years?)
    • Compensation amount ($50 Amazon card? Different?)
    • Distress referral resources (who to contact if a participant gets upset)

Week 3–4: Finish Paper 1

  • Results + Discussion (fill in real metrics once Vartan runs generate_eval_results.py)
  • Conclusion
  • Internal review (Vartan, advisor if you have one)
  • Polish + proofread

Week 5: Submit Paper 1

  • Target: Frontiers in AI
  • Parallel: Dashboard deployment (guide in deployment/README.md)

Key Coordination Points

Vartan → You (Data Flow)

  • Week 3: Vartan commits data/indian_foods_full.csv (800+ foods)
  • Week 3: Vartan runs python paper/tables/generate_eval_results.py → real metrics
  • Week 3: You pull the metrics, fill in Paper 1 Results section

Vartan → You (Paper Feedback)

  • End of Week 4: Vartan drafts Introduction + Related Work
  • Week 5: You review his drafts, integrate with your sections
  • Week 5: Final polish before Frontiers submission

You → Vartan (IRB Data)

  • End of Week 2: Forward IRB approval requirements to Vartan
  • End of Week 2: Confirm CITI training (both of you need certificates before IRB submission)

GitHub Projects Update

Quick bash to organize the board:

# On main branch, update the Kanban board at github.com/mks257/culrag/projects
# Add new columns if needed:
# - Backlog
# - Week 1–2 (Paper 1 writing)
# - Week 3–4 (Data integration + Paper 2 IRB)
# - Week 5–6 (Submit + Deploy)
# - Week 7–8 (Recruitment + Deployment)
# - Week 8–11 (User Study)
# - Week 12 (Paper 2 writing)

Move cards:


✅ Phase 1, Phase 2, Phase 3A, Phase 3B → Done
🔄 Paper 1 Abstract, Intro, Related Work → Week 1–2
🔄 Data IFCT 2017 integration → Week 1–3 (Vartan)
🔄 IRB review + submission → Week 1–2
🔄 Frontiers submission → Week 5–6
🔄 Dashboard deployment → Week 5–7
🔄 User recruitment → Week 7–8
🔄 Pilot study → Week 8–11
🔄 Paper 2 writing → Week 11–12

Next 48 Hours

  1. ✅ Send Vartan task list (paste above)
  2. ✅ Email UNT IRB (confirm if approval is needed)
  3. ✅ Update GitHub Projects (move cards to current sprint)
  4. 🔄 Start Paper 1 Abstract (expand outline → prose)

One Thing to Flag

Dashboard deployment (Week 5–7) depends on:

  • Railway backend + OPENAI_API_KEY (or Claude API key if you switch)
  • Vercel frontend
  • feedback.json storage (needs persistent volume in production — critical for user study data)

The deployment guide is in deployment/README.md. Test it locally first (should work end-to-end now), then move to staging Week 5. This is not blocking Paper 1 submission, but it is blocking recruitment Week 8, so don't defer.

Ready to send the Vartan list? I'll hold here and you can confirm when it's out.

Clone this wiki locally