Research methodology
Simple preregistered baseline rules, completed-session inputs, next-session execution and separate chronological partitions. Synthetic tests establish implementation behavior rather than profitability.
QUANTITATIVE RESEARCH / 09
Recruiter Showcase
Verified data contracts. Recorded state. Explicit boundaries.
DOCUMENTED ENGINEERING · No historical performance or investment claim.DOCUMENTATION
RESEARCH ENGINEERING / AUDITABLE SIMULATION
From a strategy rule to an explainable financial record.
QuantLab connects deterministic Python research components, an offline simulation journal and a typed web interface. Its focus is causal decisions, precise accounting and evidence that can be inspected.
Explore the strategy contracts →Architecture at a glance
READ ONLY BY DESIGN- Invented observationsExplicit modeled time and data availability
- Python decisions & ledgerWhole shares, costs, dividends and cash identities
- SQLite journalAtomic records, idempotency and deterministic replay
- Verified adapterRead-only transaction, typed Decimal strings
- Next.js interfaceValidated contracts, accessible inspection screens
Offline simulation
Explicit incoming invented observations drive persisted decisions and simulated fills. Pause/resume, restart and duplicate-event handling are part of the engine; the dashboard only reads.
Portfolio accounting
Cash, whole-share positions, cost basis and realized/unrealized profit remain distinct. Dividend entitlement becomes a receivable before eligible cash release; Decimal identities reconcile the account.
Journal integrity
Schema checks, hash linkage and complete replay detect inconsistent records. Internal consistency is stated honestly: no external signature or rollback-proof anchor is claimed.
Typed full-stack boundary
Python is the financial authority. Pydantic and Zod validate the data contract; TypeScript formats exact money strings and plots genuine recorded observations without recalculating balances.
Engineering verification
Offline behavioral tests cover financial boundaries and persistence. GitHub Actions checks Windows and Ubuntu, with frontend lint, TypeScript and production-build checks. No invented live test counter is displayed.
Demonstration and hosting boundaries
NO INVESTMENT CLAIMThe local dashboard displays invented synthetic prices and distributions. It has no broker, live feed, arbitrary file picker, trading controls or execution endpoint. Historical returns, strategy rankings and deployable investment claims are intentionally absent.
This explanatory showcase works without Python or SQLite. Public financial demonstrations remain unavailable until a separately reviewed sanitized synthetic snapshot is explicitly approved. A future public Next.js host can serve that approved immutable artifact; a private Python service remains separate and access-controlled.
Read the evidence and limitations →