A domain-specific AI assistant built with the Databricks Mosaic AI Agent Framework to help portfolio managers make informed stock investment decisions using real-time data, user preferences, and historical market trends.
Inspired by:
➡️ Building an Investment Assistant with the Databricks Mosaic AI Agent Framework
- 🔍 Real-Time Stock Data: Uses Yahoo Finance API to fetch current stock info.
- 🧠 LLM Agent for Investment Decisions: Combines financial data, user risk profiles, and natural language prompts to produce Buy/Sell/Hold recommendations.
- 📊 Delta Table Integration: Historical prices and synthetic customer preference data loaded from Unity Catalog Delta tables.
- 📈 Agent Evaluation: Tracks cost, latency, grounding quality, and approval through Mosaic AI's built-in evaluation and review tools.
- 🔁 Human-in-the-Loop: Integrated with Review App for gathering human feedback and improving model output quality.
- 🏗 Production-Ready: Deployed via Databricks Model Serving with full observability and governance.
[User Prompt] ──▶ [LLM Agent]
│
├──▶ [Yahoo Finance API]
├──▶ [Historical Data via Delta Table]
└──▶ [Customer Preferences via Delta Table]
↓
[Recommendation + Rationale]
↓
[Trace logged to MLflow] ──▶ [Review App] ──▶ [Evaluation + Fine-tuning]
- 🔍 Real-Time Stock Data: Integrated with Yahoo Finance API for up-to-date market data
- 🧠 LLM-Driven Decision Engine: Contextual Buy/Sell/Hold decisions based on user profiles and historical trends
- 📊 Delta Tables + Unity Catalog: Used for storing and querying synthetic investor data and historical prices
- 🔁 Human-in-the-Loop Feedback: Review App integration to evaluate and improve agent outputs
- ✅ Agent Evaluation Metrics: Track latency, cost, grounding quality, and approval
- 🏗 Production-Ready Deployment: Model deployed using Databricks Model Serving with governance and observability
- Portfolio Managers and Financial Analysts
- GenAI Practitioners building LLM-based decision tools
- Databricks and Mosaic AI developers
- Anyone interested in finance + AI