CaliComp

AI-Powered Cashflow Intelligence & Decision Engine

CaliComp is a full-stack financial intelligence system that transforms raw financial signals into optimized, explainable decisions under uncertainty. It is built for individuals and businesses managing cash constraints — going beyond visualization to actively prescribe what to do next.

Live Demo: (https://caliicomp.netlify.app/login)

---

What Makes CaliComp Different

Most financial tools describe the past. CaliComp prescribes the future.

Typical Tools	CaliComp
Static or mock data	Real-time Gmail ingestion via OAuth
Simple rule-based heuristics	Constrained optimization with linear programming
Insights without actions	Actionable decisions with automated email generation
Black-box outputs	Explainable, auditable, confidence-scored results

---

System Architecture

External Data Sources (Gmail / PDFs / Receipts)
                |
        Data Ingestion Layer
                |
     Parsing & Structuring Engine
                |
   Forecasting Layer (ML + Features)
                |
      Liquidity Runway Simulation
                |
 Optimization Engine (Linear Programming)
                |
 Decision + Confidence Layer
                |
 Explainability + Action Generation
                |
        Frontend Visualization

---

Core Components

Gmail Data Ingestion

• Google OAuth 2.0 authentication for secure inbox access
• Handles text/plain and text/html (with HTML stripping)
• Extracts INR/Rs/₹ amounts, debit/credit classification, and transaction dates
• Production-grade ingestion — not mock data

Cash Runway Engine

Simulates future liquidity by projecting daily balances, detecting critical cash-out points, and estimating time-to-zero.

{
  "days_to_zero": 2,
  "critical_date": "2026-03-27",
  "daily_balances": [...]
}

Optimization-Based Prioritization Engine

Formulated as a constrained linear programming problem using PuLP:

Maximize financial stability under limited cash

Decision factors: urgency, penalty severity, flexibility, liquidity impact, revenue blocking, credit impact, grace period, and penalty growth rate.

Decision Output Layer

Produces selected vs. deferred payment sets, priority rankings, and a full scoring matrix per obligation.

Confidence Scoring System

Scores are normalized between 0 and 1 based on inter-option score gaps:

Level	Meaning
High	Strong, clear decision
Medium	Moderate certainty
Low	Ambiguous — review recommended

Explainability Engine

Every decision surfaces a human-readable rationale, for example:

"Selected due to high penalty, low flexibility, and near-term due date."

All outputs are deterministic and auditable.

AI Email Action Generator

Context-aware vendor/bank/employee email drafts generated per decision:

• Vendor communications — polite tone
• Bank communications — formal tone
• Employee communications — transparent tone

Scenario-Based Planning

Simulates multiple financial futures in parallel:

• Normal trajectory
• High-revenue (e.g. festival or seasonal spike)
• Worst-case

ML Forecasting Layer

Lightweight models for revenue forecasting:

• Features: day of week, holiday flag, season, historical revenue
• Models: linear regression, moving average

Edge Case Handling

Gracefully handles empty inputs, zero cash, negative balances, missing fields, and parsing failures throughout the pipeline.

---

Tech Stack

Backend

• FastAPI, Python
• PuLP (Linear Programming)
• Gmail API (OAuth 2.0)
• Regex parsing, MIME decoding, HTML cleaning

ML Layer

• Linear regression
• Feature engineering

Frontend

• React, Tailwind CSS
• Recharts / D3

---

API Reference

Endpoint	Description
POST /api/email-ingest	Fetch and parse Gmail transactions
POST /api/runway	Simulate liquidity runway
POST /api/prioritize	Run optimization and return payment decisions

---

Status

• Real data ingestion (Gmail OAuth)
• Optimization engine (linear programming)
• ML forecasting integration
• Explainable decision outputs
• End-to-end pipeline
• Frontend visualization

---

Design Philosophy

Decision over visualization — actionable intelligence, not passive dashboards.

Explainability over black-box — every output is transparent and interpretable.

Real data over mocks — built to operate on actual financial inputs from day one.

---

Impact

• Helps startups monitor and manage burn rate in real time
• Enables smarter, data-driven payment prioritization
• Surfaces liquidity crises before they become critical