Problem Statement
Flood budgets hide across programs — locally funded, foreign-assisted, “other infra” — exactly what DPWH now promises to fix for 2027. Citizens hear “₱545B since 2022” but can’t answer for their river: how much, who got it, was it finished? Reporters trace contractor-politician links by hand across PDFs.
Proposed Solution
Search city or contractor: BudgetBaha shows projects, amounts, status, and contractor network (shared addresses, repeat winners) with red flags (single bidder, repeat delays, cost-per-meter outlier). LLM writes “Sa Malabon: ₱210M sa 6 projects, 2 unfinished, 1 contractor nanalo ng 4” with source links and a one-click FOI draft.
Target Market
- Primary users: Youth watchdogs, campus journalists, flood-hit residents.
- B2G: COA auditors, DILG, reformist LGUs.
How AI Is Used
- Document reader: LLM extracts project rows from budget PDFs and news archives.
- Red-flag scorer: Rules + anomaly detection on bidding patterns.
- Network mapper: Entity resolution links contractors, officials, addresses.
PLDT / Smart / ePLDT / Rokid Angle
- Smart: SMS budget query (“BUDGET Malabon” returns totals).
- ePLDT: Hosts budget graph for media/LGU war rooms.
- PLDT Enterprise: Transparency bundle for provinces.
Demo Scenario (3 minutes for judges)
- Search mock “Malabon”; totals + map appear with 2 unfinished flags.
- Click contractor; network shows 4 wins + shared address.
- FOI draft generates with exact project IDs.
- Compare to 2027 ₱103B flood plan view.
- Close: “Bawat piso, may pinagpunan.”
30-Hour Build Plan
| Hours | Task | Role |
|---|---|---|
| 0–3 | Mock budget rows + FOI template | All |
| 3–12 | Reader + scorer + network | AI lead |
| 3–10 | Search + SMS | Backend |
| 10–20 | Graph + explainer UI | Frontend |
| 20–26 | Accuracy review | All |
| 26–30 | Pitch, disclosure | PM / Pitch |
Suggested stack: Next.js + network graphs, FastAPI, LLM budget explainer, Postgres. Public data / APIs: DPWH budgets, COA reports, PhilGEPS (mock for demo), news trial coverage.
Risks & Privacy Notes
- No accusations: Flags are “for audit,” persons presumed innocent; Sandiganbayan outcomes tracked neutrally. Demo data fictionalized; no real personal data into AI tools.
- Privacy: Contractor data from public records only.
- IP: Public docs cited; team owns scorer.
AI Use Disclosure (draft)
Uses LLM document reading plus scoring/network code. Coding assistants helped with graphs. All demo flags human-checked against mock ground truth.
Rubric Scoring
| Criterion | Weight | Score (1–5) | Why |
|---|---|---|---|
| Innovation & creativity | 25% | 5 | Barangay-level flood-budget graph is powerful |
| Problem–solution fit | 25% | 5 | Core of the 2026 scandal debate |
| AI application & value | 20% | 4 | Doc + network AI essential |
| Feasibility | 15% | 4 | Mock data demoable |
| Impact & scalability | 15% | 4 | Every flood city |
| Weighted total | 4.50 |