Economic Pulse

AI-Powered Multi-Agent Platform for Transparent, Explainable Economic Intelligence

"Bringing clarity, collaboration, and confidence to economic decision-making."

Inspiration

Economic Pulse was born from the need for transparent, explainable, and collaborative economic analysis.
In a world flooded with data, decision-makers often rely on dashboards that show what happened but not why.
Traditional models lack context, explainability, and the ability to simulate real-world policy scenarios.

Economic Pulse bridges this gap — fusing AI explainability, multi-agent reasoning, and real-time data into one interactive ecosystem where insights are debated, explained, and justified.

What We Learned

Integrating SHAP explainability into economic forecasting models for transparency.
Creating persistent multi-agent personas capable of debate, consensus, and reasoning.
Orchestrating FastAPI microservices with modular endpoints for scalability.
Designing Streamlit-based dashboards that make complex analytics accessible.
Handling frequency-aware polling for real-time FRED API updates.
Building a transparent architecture where users can trace every model decision.

How We Built It

Architecture Overview

Layer	Components	Description
Data Layer	FRED API, cached polling, release calendar sync	Real-time data ingestion with frequency awareness
Backend (API)	FastAPI microservices	Forecasting, scenario testing, equity analysis, agent orchestration
Explainability	SHAP + model introspection	Feature importance and driver visualization
Agent Framework	Personality-driven reasoning engine	Multi-round debates and adaptive stances
Frontend (UI)	Streamlit modular dashboard	Tabs for agents, scenarios, explainability, and equity

Agent Roles

Agent	Role
Optimist	Highlights positive macro trends and growth potential
Pessimist	Focuses on risk, uncertainty, and economic vulnerabilities
Equity Analyst	Analyzes social and demographic fairness impacts
Data Scientist	Provides model-driven reasoning using SHAP and diagnostics
Moderator	Synthesizes debate outcomes into clear policy recommendations

Core Modules

Forecasting Engine
- Supports statistical (ETS/ARIMA), ML (LightGBM/XGBoost), and DL (N-BEATSx, NHITS, TiDE, TFT) models via Nixtla.
- Produces interpretable forecasts with confidence intervals.
Scenario Testing
- Simulate shocks such as inflation spikes, stimulus packages, or rate changes.
- Compare baseline vs scenario outcomes interactively.
Equity & Demographic Analysis
- Visualize demographic breakdowns (race, age, gender, income).
- Detect and quantify equity gaps for policy planning.
Explainability Layer
- SHAP value plots and plain-language model interpretations.
- Interactive visualization of top contributing features.
Audit & Transparency
- All API calls, agent decisions, and data transformations are traceable.
- Supports full reproducibility and accountability.

Challenges Faced

Making SHAP plots render in headless environments (FastAPI + Streamlit).
Designing long-term agent memory for personality consistency.
Implementing frequency-aware polling for multi-source data feeds.
Merging heterogeneous analysis modules (forecasting, equity, scenarios).
Balancing performance versus interpretability in a real-time environment.

The Economic Problem

Problem	Solution by Economic Pulse
Opaque and unexplainable models	Explainability via SHAP and transparent agent debate
No scenario testing tools	Dynamic simulation of economic shocks
Lack of equity focus	Demographic impact analysis and fairness evaluation
Fragmented analytics	Unified agent-driven reasoning framework

Economic Pulse empowers economists and policymakers to see not just the forecast, but the reasoning behind it.

Target Users & Scenarios

User	Use Case
Economists	Analyze indicators, identify drivers, and test interventions
Policymakers	Simulate “what-if” shocks and assess policy impacts
Financial Analysts	Interpret macro risks and track multi-agent consensus
Educators/Students	Learn forecasting, debate, and model explainability
Data Scientists	Validate models, explore SHAP insights, and audit outcomes

Features

Multi-Agent Economic Debate
Multi-round debates among Optimist, Pessimist, Equity, and Data Scientist agents with moderator synthesis.
Scenario Simulation
Simulate inflation spikes, recessions, or fiscal stimulus and compare against baseline.
Equity & Demographic Insights
Measure policy impact on different demographic segments.
Forecast Explainability
SHAP-driven visuals and natural language explanations of feature influence.
Transparent Tool Trace
Every agent action and API call is logged for full auditability.
Policy Recommendation Engine
Synthesized insights from agent consensus into actionable policy suggestions.

Future Enhancements

1. Broader Data Integration

Connect APIs from BLS, World Bank, OECD, IMF, Quandl, and more.
Include financial indicators such as Treasury yields, oil prices, and jobless claims.
Support user-provided CSV or Google Sheets data uploads.

2. Real-Time Adaptivity

Add WebSocket streaming for instant updates.
Implement FRED /series/updates endpoint for release tracking.
Push notifications via Slack or Email when new data is published.

3. Advanced Scenario & Equity Modules

Causal inference for policy impact estimation.
Debate-driven model selection among agents.
Multi-dimensional equity diagnostics.

4. Personalized Experience

User risk profiles (Conservative, Balanced, Aggressive).
Adaptive tone and recommendations based on preference memory.

5. Forecast Fusion

Combine outputs from StatsForecast, MLForecast, and NeuralForecast models.
Coherence checks and uncertainty comparisons across methods.
Agent reasoning on model disagreement.

6. Gamified Forecast Challenge

Users compete in “Forecast Challenge Mode” using RMSE/CRPS scoring.
Educational version for universities and policy schools.

7. Policy Optimization

Integrate PuLP or OR-Tools for policy mix optimization.
Equity-constrained simulations for fair allocation of resources.
Visual trade-offs between cost, impact, and fairness.

Team Information

Team Members

Sai Krishna Jasti – AI Architect & Backend Developer
Ankit Hemant Lade – Data Scientist & Economic Intelligence Lead

Contact Emails

Vision

Economic Pulse is more than a dashboard — it is a living ecosystem for economic intelligence.
By uniting explainable AI, multi-agent reasoning, and policy transparency, it enables data-driven democracy in economic planning.