Originally published on my personal engineering blog at louvincentbaroro.me.
At the ASEAN AI Hackathon 2026, our team Syntaxure SEA set out to solve a massive problem in Southeast Asia: Environmental accountability.
Citizens witness environmental violations every day. Illegal logging, open burning, and hazardous waste are rampant. But reporting is fragmented, and whistleblowing is dangerous.
To solve this, we built LikasLens, a neuro-symbolic civic intelligence platform that allows citizens to snap photos of violations, dynamically routes them to the correct government agency using AI, and protects the whistleblower's identity.
Advanced Platform Features
LikasLens is far more than just a reporting tool. Some of the most complex features we engineered include:
- **Eco-Credit Rewards System: Users earn points for verified reports that can be redeemed at major partner stores (e.g., proposed integrations with SM, Jollibee, Globe).
- Report Chaining & Anti-Sybil Logic: We implemented GPS-proximity clustering. If multiple users report the same issue within a 50m radius, the system corroborates and auto-escalates it.
- Pattern Escalation: The platform detects spatial-temporal patterns to trigger urgent responses for repeated violations in the same barangay.
- Neo4j GraphRAG Integration: The system auto-detects the exact jurisdiction and law violated based on the user's country using AI.
The Engineering Architecture
Building a platform of this magnitude required extreme focus. My partner Jeff Martinez and I split the architecture perfectly down the middle.
I was responsible for architecting the entire user-facing structure and interactive systems using a Next.js 16 Monorepo. My core focus included:
- Interactive Data Visualization: Engineering MapLibre GL heatmaps and 3D interactive globes (cobe.js) to visualize violation hotspots.
- Blockchain UI: Building the frontend for SHA-256 evidence hashing to ensure a tamper-proof chain of custody.
-
Multi-Language Structure: Laying out the
next-intllocalization structure for 10 ASEAN languages.
Meanwhile, Jeff engineered the Laravel 12 backend and Python/FastAPI microservices, integrating dual YOLOv8 vision models and ONNX Runtime WebAssembly directly into the mobile flow.
The Results
Out of hundreds of teams, we finished as Semi-Finalists (Top 40 Overall) and placed in the Top 10 for the Climate Track.
While we barely missed the grand prize (a flight to Vietnam), we walked away with a piece of engineering that proves exactly what we are capable of. We built a massive, enterprise-grade neuro-symbolic platform capable of handling multi-language support, blockchain hashing, and jurisdiction-aware AI routing.
If you want to read more of my engineering notes or check out my other AI projects, visit my portfolio at louvincentbaroro.me.
Top comments (1)
Neuro-symbolic feels like exactly the right fit for civic reporting, because you need perception (classifying the photo of a violation) and hard, explainable rules (does this actually breach a regulation, and which one). How did you split responsibility between the two layers - is the symbolic side encoding the regulatory logic, with the neural side just doing extraction and classification? Curious whether the symbolic layer also gave you the auditability that matters so much for anything a citizen might report.