Extracted from First Meet slides + Design Thinking workshop. Team: SinkAlert (BD002, Safety Track). Updated: 16 Jun 2026.
| Real gaps in AI — Thai industry | ✅ |
| Thai LLM + Agentic AI | ❌ |
| Community tools / B2B or B2C | ✅ B2G+B2C |
| Crowdsource / cross-sector data | ✅ 4 sources |
| Working PoC, deployed on cloud | ✅ Live |
| Actually runs (not just mockup) | ✅ Running |
| Verified demand (sign-ups, etc) | ❌ 0 validated |
| Profitable / social value | ✅ Clear |
| Rule | Status | Note |
|---|---|---|
| Attribute all source licences | ✅ | All data sources documented in research |
| Adapt to Thai context, contribute original idea | ✅ | Bangkok-specific sinkhole prediction |
| List licences and provenance of pre-existing code | ⚠️ | Need to compile licence list |
| >50% functional logic created during hackathon | ⚠️ | Models built pre-hackathon — need to verify ratio |
| No plagiarism / fabrication / misuse | ✅ | All research cites sources |
| Resource | Status | Action |
|---|---|---|
| AWS Credits (~$1,400/team) | ❌ | Claim at Kick-off 19 Jun |
| NVIDIA (build.nvidia.com) | ❌ | Explore GPU compute for training |
| Google Cloud Credits | ❌ | Claim at Kick-off |
| ThaiLLM Access | ❌ CRITICAL | Integrate analysis/recommendation layer |
| BDI Curated Open Data | ⚠️ | Request at Kick-off, integrate |
| Mentors | 🔜 | Come with questions ready |
| # | Gap | Judging Criterion Affected | Deadline |
|---|---|---|---|
| 1 | No user validation — 0 sign-ups, 0 surveys, 0 interviews | SHIP REAL #3: Verified demand | ASAP |
| 2 | No ThaiLLM integration | EXPLORE GAPS #2 | 26 Jun |
| 3 | No agentic AI demonstration | EXPLORE GAPS #2 | 26 Jun |
| 4 | >50% originality unverified | Rules: Novelty assessment | 10 Jul |
| Phase | Status | Evidence |
|---|---|---|
| Empathize Users: citizens, BMA engineers, insurers, construction firms | ❌ | No interviews conducted yet |
| Define "How might we provide affordable city-wide sinkhole early warning?" | ✅ | Clear problem statement articulated |
| Ideate Tiered screening, 3D viz, LINE Bot, agentic AI fusion | ✅ | Multiple solutions generated |
| Prototype 3D Twin Map, API, XGBoost pipeline, dashboard | ⚠️ | Built but model not trained |
| Test User feedback, model accuracy, response time | ❌ | No testing conducted |
📄 Full analysis: GitHub docs/004 | 📄 Raw extraction: GitHub docs/003