BYLD Wealth — byld-portfolio Service Blueprint

Domain Model

Portfolio Aggregate

«Aggregate Root» Portfolio

- id : PortfolioId PK

- userId : UserId

- holdings : List<Holding>

- totalValue : Money

- investedAmount : Money

- xirr : Percentage

- lastSyncAt : Instant

+ addHolding(holding) : void

+ recordTransaction(txn) : void

+ computeXIRR() : Percentage

+ syncFromAA(data) : void

«Entity» Holding

- id : HoldingId

- isin : ISIN

- fundName : String

- units : BigDecimal

- avgNav : Money

- currentNav : Money

- currentValue : Money

- category : AssetCategory

- source : DataSource

+ unrealizedGain() : Money

+ gainPercent() : Percentage

«Entity» Transaction

- id : TransactionId

- holdingId : HoldingId

- type : TxnType {BUY, SELL, SIP, SWITCH, DIV}

- amount : Money

- units : BigDecimal

- nav : Money

- executedAt : Instant

- status : TxnStatus

«Value Object» Money

- paise : long

- currency : Currency = INR

Value Object Deep Dive

Money — "long paise" Representation

Money

All financial calculations use "long paise" internally. No floating point. No rounding errors. Ever.

Internal: 1_24_56_78_90 paise

Type: long (64-bit signed)

Max: 92,23,37,20,36,854.76

Currency: INR (ISO 4217)

Display Format

₹12,45,678.90

Indian numbering system (lakh, crore)

add(other: Money)

50,000.00 + 25,000.50

= 75,000.50

subtract(other: Money)

1,00,000.00 - 12,345.67

= 87,654.33

multiply(factor: BigDecimal)

1,00,000.00 * 1.12

= 1,12,000.00

allocate(ratios: int[])

1,00,000.00 / [60, 30, 10]

= [60K, 30K, 10K] (no loss)

isGreaterThan(other)

50,000.01 > 50,000.00

= true

display(locale: IN)

124567890 paise

= "₹12,45,678.90"

Morningstar NAV Sync

Scheduler

Morningstar

Transform

Daily 23:00 IST | ~5000 schemes | bulk upsert

Account Aggregator Import

User

Finvu

Transform

Holdings

Consent → FI Fetch → ReBIT v2 → normalize → merge

CQRS Architecture

Write Model → Events → Read Model

Write Model

PostgreSQL

portfolio schema
Normalized 3NF
Strong consistency

Kafka

Event Store

Kafka + MSK

PortfolioUpdated
HoldingChanged
NAVRefreshed

Consumer

Read Model

Redis Cache

Dashboard JSON
Pre-computed
TTL: 60s

Performance Analytics Engine

+18.7%

XIRR

Annualized return

+15.2%

CAGR

3-year compound

1.12

Sharpe

Risk-adjusted

+3.2%

Alpha

vs Nifty 50 TRI

0.87

Beta

Market sensitivity

-8.4%

Max DD

Max drawdown

Benchmark Comparison

Your Portfolio +18.7%

Nifty 50 TRI +15.5%

CRISIL Hybrid +12.1%

FD (SBI 1Y) +7.1%

REST API Endpoints

	Endpoint	Description	Notes
GET	/api/v1/portfolio	Dashboard summary	Redis read model
GET	/api/v1/portfolio/holdings	All holdings with current value	Paginated
GET	/api/v1/portfolio/holdings/{id}/transactions	Transaction history for holding	Cursor-based
GET	/api/v1/portfolio/performance	XIRR, CAGR, benchmark comparison	Computed daily
POST	/api/v1/portfolio/import/aa	Trigger AA import via Finvu	Async, consent req
POST	/api/v1/portfolio/sync	Force Morningstar NAV refresh	Rate limited

Domain Events (Kafka)

PortfolioCreated HoldingAdded HoldingRemoved TransactionRecorded NAVRefreshed AAImportCompleted XIRRComputed DashboardCacheInvalidated

Topic: byld.portfolio.events
Partitions: 6 (by userId hash)
Retention: 7 days
Consumer groups: advisory, mia, dashboard

Architect's Notes

Money as "long paise" is non-negotiable. BigDecimal is tempting but slow. IEEE 754 double is unacceptable for financial math. The allocate() method handles Foemmel's Conundrum (splitting ₹1.00 three ways without losing a paisa).

CQRS separates the write path (transactions, syncs) from the read path (dashboard, charts). Redis read model is rebuilt from events -- eventually consistent but renders the dashboard in <50ms.

"If you're doing money arithmetic with doubles,
you will lose money." -- Joshua Bloch, Effective Java

Legend

Domain / Business

Infrastructure

Events / Async

External Services

Security / Auth

Data / Storage