🤖 Interested in algorithmic trading, AI systems, and backend engineering
📈 Currently building crypto futures simulation & AI quant trading platforms
🚀 Deploying real systems — not just notebooks
📍 From Indonesia
AI quant trading platform · Backtesting · Multi-agent LLM · 10+ exchange support · Docker + Vue.js
Automated trading bot · Real-time market data · Signal execution pipeline
Futures price prediction · ML models · Feature engineering
AgriTwin 🌱
AI greenhouse digital twin platform — production-grade evolution from 17K-line Streamlit monolith
📖 AgriTwin Details — click to expand
AgriTwin is a full-stack AI-powered digital twin platform for greenhouse management, evolved from a single-file Streamlit prototype (tumbal.py, 15,600+ lines) into a distributed three-tier architecture through a structured 5-phase migration — with zero downtime.
| Layer | Technology | Notes |
|---|---|---|
| Frontend | Next.js 15, React 19, Tailwind CSS | Dark-mode dashboard, mobile-responsive |
| Backend | FastAPI (Python), async, WebSocket | 9 REST endpoints + 1 WebSocket |
| Database | Supabase (PostgreSQL + Realtime) | 7 tables, dual-write with SQLite fallback |
| IoT | HiveMQ Cloud (MQTT TLS) | ESP32 → cloud → dashboard in <2s |
| Weather | Open-Meteo (free, no API key) | Current + 16-day forecast + 7-day history |
| AI | Gemini 2.0 Flash + RAG | 10 Indonesian agronomy documents as knowledge base |
| Monitoring | Sentry + PostHog | Error tracking + product analytics |
| Deploy | Vercel (FE) + Railway (BE) + Docker | CI/CD via GitHub Actions |
- 🌡️ Real-time IoT sensor monitoring — temperature, humidity, CO2, soil moisture, pH, EC
- 🌦️ Live weather integration — Open-Meteo primary, OpenWeatherMap fallback, physics-based simulation as last resort
- 🤖 AI Agronomist — multi-LLM (Gemini/Groq/Ollama) with RAG context injection from local agronomy knowledge base
- 📸 Plant Doctor — disease detection from photos via Gemini Vision
- 🚨 Alert Engine — 9 threshold rules, auto-evaluation, Telegram notifications
- 💰 Market prices — 23 Indonesian crops, sourced from PIHPS BI / World Bank / BPS
- 🏔️ 3D terrain visualization — real SRTM DEM from AWS Terrarium tiles + hypsometric colorscale
- 🧬 Genetic optimizer — NSGA-II multi-objective setpoint optimization
- 📊 Economics engine — full P&L, ROI, payback period, carbon credit MRV
- 🔗 Dual-write persistence — SQLite local + Supabase cloud (graceful degradation)
GET /api/health → service status (Supabase, MQTT, weather)
GET /api/weather/{lat}/{lon} → current weather + forecast
POST /api/sensors/ingest → receive ESP32/client sensor data
GET /api/sensors/{zone_id} → sensor history
GET /api/alerts → active alerts
POST /api/alerts/{id}/acknowledge
GET /api/market/prices → commodity prices (23 crops)
POST /api/ai/query → AI agronomist (Gemini + RAG)
WS /ws/zones/{zone_id}/live → real-time sensor WebSocket
| Phase | Scope | Status |
|---|---|---|
| 0 | Security — .env, dotenv, Sentry, .gitignore |
✅ |
| 1 | Data Layer — Open-Meteo, Supabase, market prices | ✅ |
| 2 | IoT Real — HiveMQ MQTT, alert engine, ESP32 spec | ✅ |
| 3 | Backend Split — FastAPI + Next.js + Streamlit legacy | ✅ |
| 4 | Production — Docker, CI/CD, RAG, deploy configs | ✅ |
Python FastAPI Next.js React TypeScript Tailwind CSS PostgreSQL Supabase Redis MQTT HiveMQ Docker Sentry PostHog Gemini API WebSocket GitHub Actions
⚡ Markets are data problems. Solve them with code.