Azure-Based Investment Management Platforms for Multi-Asset Portfolios

Summary

We build investment management platforms for portfolios that span several asset classes, custodians and legal entities, where a spreadsheet stops being a safe book of record. This kind of system maintains positions, transactions and cash, plans allocations against a mandate, values holdings on a defined schedule, measures performance, and publishes role-appropriate views to investment teams, accountants and investors. The engineering weight sits behind the screens, in producing one consistent valuation of a heterogeneous portfolio from data that never arrives in one shape.

The Challenge

Data reaches the platform from custodians, brokers and fund administrators in different formats, on different calendars, and with corrections that land after figures have already been published. A valuation that cannot be explained is worse than no valuation, so every computed number has to be traceable to the inputs that produced it. Platforms designed only around listed instruments break the moment private holdings, commitments, capital calls and irregular valuations appear. Performance methodology buried in code becomes impossible to defend. Where one deployment serves several organisations, an entitlement mistake exposes one investor's holdings to another, which makes the isolation model an early architectural decision rather than a configuration detail.

The Solution

Service decomposition

The architecture separates concerns that evolve and scale independently, rather than one application that does everything:

• Instrument and counterparty reference data - a security master covering listed instruments, funds, private holdings, real assets and cash, with identifiers, classifications and corporate-action history.
• Transaction and position ledger - an immutable record of trades, subscriptions, capital calls, distributions, fees and cash movements, from which positions are derived rather than stored as loose balances.
• Valuation and performance engine - pricing rules per asset type, FX translation into base currency, accruals, and time-weighted and money-weighted returns gross and net of fees.
• Planning and rebalancing - target allocation models, drift monitoring, scenario comparison and proposal workflows.
• Reporting and document generation - scheduled statements and factsheets rendered from the same figures the screens show.
• Identity and entitlements - separate trust boundaries for staff and investors.

Ingestion and reconciliation

Feeds arrive as SWIFT MT535 and MT940 messages, fixed-format extracts, API pulls and, unavoidably, PDF statements that need parsing and human confirmation. Market data supplies prices, FX rates, benchmarks and corporate actions. Every feed passes a reconciliation stage that compares custodian positions and cash against the internal ledger, quarantines breaks, and lets an operator resolve them with an audit trail instead of editing records directly. Downstream, the platform exports to accounting and general ledger systems and pushes documents to investor portals.

Isolation and control

Schema-per-tenant is simpler to restore selectively; shared tables with row-level security cost less to operate but raise the blast radius of an entitlement error. Either way, entitlements reach down to individual portfolios and legal entities, supported by encryption at rest and in transit, an append-only log of who saw and changed what, residency rules that may pin storage to a region, and retention driven by financial record-keeping obligations.

Sequence of work

Structured sessions with wealth managers, analysts and accountants establish how instruments, mandates and fee structures actually work, and a functional specification and domain model precede interface work. Wireframes are produced per role, because a portfolio manager, an accountant and an investor need different densities of information. Build order starts with ingestion, the ledger and reconciliation, since nothing above them is trustworthy until they are correct. Valuation logic is covered by regression tests against agreed reference datasets, and corrections are handled bitemporally: what was known then, and what is known now.

What This Delivers

The result is a single defensible view of a mixed portfolio, where a restated figure can be explained from its inputs rather than argued about. Investment staff plan allocations against live drift, accountants receive exports that agree with the screens, and investors see only their own holdings through a separate identity boundary. Because analytical queries are kept off the transactional database, reporting does not degrade daily operations.

Technologies and Tools

• Azure App Service or AKS, Service Bus for queued ingestion, Blob Storage for documents, Key Vault for secrets, and Entra ID with an external tenant separating investor from employee identity
• A .NET service layer with an Angular front end, and grid and charting libraries such as SyncFusion for dense analytical screens
• PostgreSQL for the transactional model and TimescaleDB for price, FX and valuation history