# JKoDe

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [Januda J Kodithuwakku](https://cerebralvalley.ai/u/Januda)
- **GitHub:** https://github.com/kjanuda/Jkode-GuardianX
- **Demo video:** https://www.youtube.com/watch?v=g03_YNTv9So
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2300

Guardian X — AI-Powered Telecom Network Intelligence & Customer Protection Platform

Guardian X is an AI-assisted telecom intelligence prototype designed to help engineers investigate connectivity problems, assess customer impact, forecast congestion and evaluate safer network optimization strategies.

Telecom incidents may involve cell congestion, device limitations, weak radio signals, backhaul degradation, weather or terrain. Guardian X combines these evidence sources into a single, explainable investigation workflow rather than relying on isolated alarms or unsupported AI conclusions.

Key capabilities:
1. Network Intelligence: Explore tower, sector, cell and backhaul relationships.
2. Device & Customer Analysis: Review synthetic RSRP, SINR and customer complaint scenarios.
3. Evidence-First RCA: Distinguish suspected faults, missing or contradictory evidence and healthy conditions.
4. Traffic Forecasting: Generate synthetic hourly cell-load forecasts.
5. Optimization Digital Twin: Simulate hypothetical load redistribution with RF, backhaul and handover safety checks.
6. Geography: Process real Copernicus GLO-30 DSM and ESA WorldCover data for terrain and land-cover screening; OSM building integration remains under verification.
7. Weather & Backhaul: Assess synthetic rain, microwave signal and packet-loss evidence.
8. Incidents & Alerts: Demonstrate investigation and human review workflows.

Meta AI integration:
Guardian X includes an optional Meta Model API adapter for sending minimized, structured synthetic evidence and validating candidate root-cause responses. It has passed offline tests with mocked requests and responses. A Meta Llama 3.2 benchmark workflow has also been prepared alongside an existing local Ollama investigation baseline. Live Meta hosted inference and direct Llama benchmarking have not yet been verified. AI hypotheses cannot authorize network changes.

Technology: Python, FastAPI, PostgreSQL, React/Next.js, modular AI adapters, Rasterio and geospatial datasets.

Results: 558 selected offline regression tests passed through Q61D2, plus 24 additional local FastAPI API tests. A local demo API started successfully. Real elevation and land-cover datasets were processed, while network, device, customer, traffic and weather scenarios remain synthetic.

Guardian X is a working analytical prototype and interactive offline showcase, not a production operator-connected system. Geographic findings are screening-only; field calibration, live operator integration and model validation are next steps.

Goal: make telecom troubleshooting more explainable, reliable and safe, while supporting better customer experiences.

## More from Meta Global AI Developer Hackathon

- [Measure Once](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2297)
- [Caffeinated coder](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2298)
- [Comimo](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2299)
- [GridPulse](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2301)
- [Gözüm](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2302)
- [Tessa](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2303)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2300. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt