AI-Powered Open Finance Data Platform

Normalize mock bank, account, balance, and transaction data into a common financial data model, then run explainable insight workflows across cash flow, recurring payments, and financial health signals.

4Data sources
7CDM entities
6Insight tools

Project overview

AI-powered open finance intelligence platform

This project models how modern open-finance platforms transform fragmented account and transaction feeds into normalized, application-ready intelligence. The emphasis is not a generic budgeting chatbot — it is the reusable data platform and workflow layer that makes financial insights explainable, auditable, and safe for downstream product experiences.

🏦

Financial connectivity

Mock Plaid-style institutions, accounts, balances, and transactions enter the platform as fragmented source records from multiple providers and schemas.

🧩

Open finance CDM

Source records are normalized into reusable account, institution, merchant, transaction, category, cash-flow, and signal entities.

📊

Deterministic insight tools

Recurring payment detection, cash-flow analysis, runway estimation, merchant enrichment, and affordability signals are generated through explainable services.

🤖

AI workflow orchestration

An agent workflow plans tool execution, streams workflow progress, surfaces evidence, and generates review-oriented recommendations.

Explainable recommendation layer

Signals are surfaced as evidence-backed risks and opportunities instead of opaque chatbot-style financial advice.

Problem

Financial data arrives fragmented and difficult to operationalize

Financial data providers expose accounts, balances, merchants, and transactions through inconsistent schemas, naming conventions, category systems, and timing patterns. Product teams often inherit raw transactional data without the operational context needed to build reliable financial experiences.

Raw transactions can show what happened, but they rarely explain which patterns matter, which recommendations should be trusted, or what workflow action should happen next.

Solution

Normalize first, reason second

The platform ingests synthetic open-finance data, maps it into a reusable common financial data model, enriches entities with derived metadata, and then runs an explainable workflow that coordinates deterministic financial analysis tools.

The output becomes a reusable recommendation layer: recurring charges, income volatility, subscription exposure, runway, affordability signals, and evidence-backed financial-health opportunities.

Open finance common data model

Reusable financial entities and explainable workflow orchestration

The architecture mirrors common data model patterns used in healthcare and enterprise data systems: preserve raw provider payloads, normalize them into stable operational entities, enrich them with derived signals, and expose a workflow layer that keeps recommendations inspectable and auditable.

01

Ingest

Load accounts, balances, institutions, and raw transactions from connected financial providers.

02

Normalize

Map provider-specific payloads into a reusable common financial data model.

03

Enrich

Attach merchant, category, recurrence, and cash-flow metadata.

04

Analyze

Calculate runway, volatility, subscription exposure, affordability, and financial-health signals.

05

Recommend

Generate explainable product recommendations with visible workflow traces and evidence.

System architecture

AI-native financial data and workflow platform

Frontend

Next.js operational dashboard with workflow streaming, recommendation panels, and explainable insight visualization.

API layer

FastAPI services expose normalized financial entities, workflow orchestration, and deterministic insight tools.

Normalization engine

Transforms fragmented provider payloads into reusable financial entities and derived operational signals.

Workflow orchestrator

Coordinates tool execution, workflow sequencing, streaming updates, and recommendation generation.

Insight tools

Recurring payment detection, merchant normalization, cash-flow analysis, category analysis, and risk scoring.

Recommendation layer

Produces explainable operational recommendations designed for review rather than autonomous financial action.