AlphaOps AI

AlphaOps AI is an end-to-end production-oriented AI system combining Data Engineering, MLOps, LLM-powered SQL Agents, and autonomous workflows.

The project demonstrates how to design, deploy, monitor, and continuously improve a machine learning system in a real-world financial context.

Architecture Overview

1️⃣ Data Engineering Layer (✔️ Done)

• Daily stock ingestion via yfinance
• PostgreSQL Star Schema (DimTickers, DimTime, FactOHLCV)
• Batch orchestration with Apache Airflow

2️⃣ LLM SQL Agent (✔️ Done)

• LangChain + Mistral Codestral
• Natural language → validated SQL (SELECT-only)
• Schema-grounded RAG
• Secure SQL validation layer

3️⃣ ML Forecasting Layer (✔️ Done)

• LSTM time-series prediction model
• Feature engineering (returns, volatility, rolling metrics)
• Model versioning
• Performance tracking

4️⃣ MLOps & Monitoring (✔️ Done)

• Model serving with FastAPI
• User dashboard via Streamlit
• Automated monitoring (performance, drift, quality) using Evidently
• Metrics storage in PostgreSQL

5️⃣ Autonomous AI Agent (✔️ Done)

Orchestrated via n8n + LLM:

• Sends automated email notifications
• Explains model predictions
• Collects user feedback
• Analyzes feedback using an LLM
• Triggers retraining when thresholds are reached

6️⃣ Infrastructure (✔️ Done)

• Fully containerized with Docker Compose
• Modular microservices architecture
• CI/CD ready structure

---

Key Objective

Build a production-grade AI system that bridges:

Data Engineering × LLM Agents × MLOps × Autonomous Decision Systems