DAIA - Deriv AI Analyst

DAIA (Deriv AI Analyst) is a real-time enterprise observability platform built for Deriv's business operations. It monitors Active Daily Users, Trading Volume, Regional Performance, Platform Health, and Instrument Activity, automatically detecting anomalies requiring executive attention. The system uses a 4-layer intelligence pipeline: a Correlation Engine for streaming telemetry, an LLM Reasoning Agent for contextualizing deviations, a Severity Scorer using z-score analysis, and an Executive Briefing Generator for actionable reports. DAIA operates through three Gemini-powered agents: Agent 1 (Analyst): Performs statistical anomaly detection across regions (NA, EU, APAC), platforms (Trader, MT5), and instruments (Synthetics, FX, Stocks) using rolling baselines, percentage deviations, and z-scores. Agent 2 (Reporter): Generates executive briefing reports with Market Overview, Regional Drivers, Risk Commentary, and prioritized Action Items with timelines. Reports are downloadable. Agent 3 (Advisor): Interactive chat interface for follow-up questions, root cause analysis, and investigation path exploration using live analysis data. The platform features a healthy-to-anomaly state transition. By default, the dashboard shows all systems nominal. When new telemetry is uploaded, the system processes it in real-time — KPI cards turn red, trend charts reveal drops, regional charts expose impacted areas, and an investigation path traces the anomaly from region to platform to instrument. Key features: live CSV upload with Gemini analysis, 28-day ADU trend chart, dynamic regional and instrument charts, auto-generated executive reports, and persistent AI chat for deep-dive investigation. The backend uses Python/Streamlit with 3-tier anomaly detection. The frontend uses React, TypeScript, Recharts, and Gemini API via Google AI Studio. DAIA transforms raw business telemetry into executive intelligence — turning data noise into actionable decisions within seconds.