Live App →

Portfolio Intelligence

Deterministic analytics, risk metrics, and report generation for comprehensive portfolio analysis and client communication.

Portfolio Intelligence

Scope — Quantitative portfolio analytics engine that computes risk-adjusted returns, concentration metrics, benchmark comparisons, and generates publication-ready reports for wealth managers and family offices.

Executive Summary

Portfolio Intelligence transforms extracted holdings data into actionable analytics. It combines deterministic financial calculations (performed in the Agentic Backend) with LLM-generated narratives (via the LLM Orchestrator) to produce comprehensive portfolio review reports, risk dashboards, and rebalancing recommendations.

The engine supports both real-time chat queries (e.g., “What’s my Sharpe ratio?”) and batch report generation (11-section PDF/PPTX portfolio review).

The Problem

Wealth managers need to answer sophisticated questions that generic portfolio tools cannot handle:

  • “What is the risk-adjusted return of my AIF allocation versus the benchmark?”
  • “How much overlap exists between my large-cap mutual funds and PMS strategies?”
  • “Simulate the impact of moving 10% from equity to gold on portfolio volatility.”

Excel-based analysis is fragile, non-collaborative, and impossible to generate on-demand during client meetings.

Architecture

┌─────────────────────────────────────────────────────────────────┐
│                    Portfolio Intelligence                        │
├─────────────────────────────────────────────────────────────────┤
│                                                                  │
│  ┌─────────────┐      ┌──────────────────┐      ┌────────────┐ │
│  │   Nexus     │─────▶│  Agentic Backend │─────▶│  Sentinel  │ │
│  │  Extraction │      │  Portfolio Engine│      │  Frontend  │ │
│  └─────────────┘      │                  │      │ (Charts)   │ │
│                       │ • PyPortfolioOpt │      └────────────┘ │
│                       │ • yfinance       │             ▲       │
│                       │ • Risk Metrics   │             │       │
│                       └────────┬─────────┘             │       │
│                                │                       │       │
│                                ▼                       │       │
│                       ┌──────────────────┐             │       │
│                       │  LLM Orchestrator│─────────────┘       │
│                       │  (Narrative Gen) │                     │
│                       └──────────────────┘                     │
└─────────────────────────────────────────────────────────────────┘

Calculation Modules

Module Library / Method Output
Risk Metrics empyrical / custom Sharpe, Sortino, VaR, CVaR, Max Drawdown
Optimization PyPortfolioOpt Efficient frontier, min-volatility, max-Sharpe
Return Analysis yfinance + XIRR Annualized, YTD, Q-o-Q, benchmark-relative
Overlap Matrix Custom intersection Fund-level and stock-level overlap scores
Scenario Engine Monte Carlo Projected returns under allocation shifts

Portfolio Review Report (11 Sections)

  1. Executive Summary
  2. Asset Allocation Overview
  3. Fund Overlap Analysis
  4. Benchmark Comparison
  5. Risk Metrics & Ratios
  6. Diversification Score
  7. Sectoral Concentration
  8. Geographic Exposure
  9. Historical Performance
  10. Scenario Analysis
  11. Recommendations Narrative

Personas & Journeys

Relationship Manager

  1. Uploads client portfolio documents to Zen chat
  2. Requests: “Generate portfolio review report”
  3. System extracts holdings, fetches live NAVs, computes metrics
  4. Receives 11-section PDF with charts and AI commentary
  5. Reviews PII-masked summary before sharing with client
  6. Schedules follow-up meeting from embedded action card

Family Office CIO

  1. Runs quarterly book-wide analysis across 50+ entities
  2. Aggregates all CAS and AIF statements via batch pipeline
  3. Reviews concentration heatmap across the family
  4. Identifies overlap: 4 entities hold overlapping mid-cap exposure
  5. Runs scenario: “Reduce equity by 10%, increase gold and international”
  6. Receives impact analysis on volatility and expected return

Risk Officer

  1. Monitors VaR and CVaR across discretionary mandates
  2. Sets alert thresholds: Max 15% sector concentration, Max 20% unlisted
  3. Receives automated weekly risk dashboard
  4. Drills into flagged clients for detailed analysis
  5. Exports compliance-ready report for SEBI filing

Key Features

Feature Detail
Live Price Feeds yfinance integration for end-of-day NAVs and stock prices
Deterministic Calculations No LLM hallucination in numbers; all metrics computed with validated formulas
Benchmark Comparison Nifty 50, Nifty Midcap 100, CRISIL Hybrid, custom benchmarks
Fund Overlap Matrix Visual matrix showing duplicate stock exposure across funds
Efficient Frontier Modern portfolio theory visualization with current portfolio position
Scenario Modeling User-defined allocation shifts with projected risk/return
PII Masking Client names and sensitive identifiers masked in shared reports
Export Formats PDF, PPTX, Excel

API Surface

Method Endpoint Purpose
POST /api/v1/orchestrator/invoke Portfolio review with intent portfolio_review
POST /api/v1/agent/invoke Direct portfolio agent invocation
GET /api/v1/dashboard Aggregated portfolio metrics
POST /api/v1/chat/invoke Chat-based portfolio query

Security, Compliance & Operations

  • Deterministic Guarantees — All quantitative metrics use validated open-source libraries; no LLM involvement in calculation logic
  • Data Freshness — Price feeds cached with 24h TTL; manual refresh available
  • Benchmark Accuracy — Indices mapped via standard tickers (^NSEI, ^CRSLDX, etc.)
  • Regulatory Alignment — SEBI IA risk disclosure norms, PMS performance reporting standards
  • Audit — Every report generation logged with input holdings hash, model versions, and parameter set