Back to projects
Active2024 — PresentEnterprise

Anti-Money Laundering (AML) Application

Real-time transaction monitoring and risk scoring for regulatory compliance teams.

Private — case study only

Period

2024 — Present

My role

Lead Architect & Full-Stack Engineer

Team

5 engineers (cross-functional)

Client

Dooit.ai

The Problem

Why this needed to exist.

Compliance officers at mid-market financial institutions review thousands of transactions daily but lack tooling to surface suspicious patterns in real time. Manual ledger digging delays Suspicious Activity Reports (SARs) and exposes the institution to regulatory penalties.

The Approach

How I solved it.

Designed an event-driven pipeline that ingests transactions in real time, scores them against a rule engine plus an ML anomaly model, and routes high-risk activity into an investigator console. Audit trails are cryptographically signed and exportable for regulators on demand.

Personas

Who I designed for.

Each persona shaped a specific surface of the product. Goals and pain points were validated through interviews and shadowing.

SO

Sarah

Compliance Officer

Goals

  • Triage flagged transactions within minutes
  • File SARs without leaving the console
  • Maintain a defensible audit trail

Pain Points

  • 70% false-positive rate from legacy systems
  • Switches between three tools to investigate one transaction
  • No replay or what-if testing for new rules
RA

Daniel

Risk Analyst

Goals

  • Author and tune risk rules safely
  • Backtest against historical data
  • Monitor precision over time

Pain Points

  • Rule changes require engineer hand-offs
  • No staging environment for what-if simulation
  • Hard to attribute alerts to specific rules
EA

Imran

External Auditor

Goals

  • Verify regulatory compliance end-to-end
  • Export evidence reports on demand
  • Trace any decision back to its rule

Pain Points

  • Reports require engineering hand-offs
  • Trace logs scattered across services
  • No standardized export format

Use Cases

Key user flows.

The most critical scenarios the product is designed to make effortless.

01

Compliance Officer

Triage a flagged transaction

  1. 1Receive real-time alert in priority queue
  2. 2Inspect transaction context: parties, history, score breakdown
  3. 3Mark as benign, escalate, or initiate SAR
  4. 4Audit trail captured automatically
02

Risk Analyst

Configure a new risk rule

  1. 1Author rule in DSL editor with autocomplete
  2. 2Backtest against last 90 days of transactions
  3. 3Promote to staging, monitor precision
  4. 4Promote to production behind feature flag
03

External Auditor

Generate a compliance report

  1. 1Select date range and jurisdictions
  2. 2Choose export format (PDF / CSV / JSON)
  3. 3Verify cryptographic signature on export
  4. 4Download with one click

UX Process

How I got from problem to product.

The end-to-end design process — from research to ship.

  1. 01

    Discovery

    Stakeholder interviews with compliance leads at three pilot banks; mapped existing workflows and friction points.

  2. 02

    Information Architecture

    Defined the entity model — Transaction, Alert, Case, Rule — and access matrix per role.

  3. 03

    Wireframes

    Low-fi flows for the top 5 use cases, validated with two compliance officers.

  4. 04

    Design System

    Built tokens, components, and a WCAG AA accessibility baseline on top of Tailwind primitives.

  5. 05

    Usability Testing

    Moderated tests with 6 users; reduced average triage time by 42% versus the legacy tool.

  6. 06

    Iteration & Ship

    Phased rollout behind feature flags with observability stack from day one.

Roadmap

What shipped — and what's next.

Phased rollout, with each phase validating learnings from the last.

  1. Phase 1 — Core monitoring engine

    Shipped

    Q1 2024

    • Real-time transaction ingestion pipeline
    • Rule engine + scoring service
    • Investigator console MVP

  2. Phase 2 — Workflow & SAR filing

    Shipped

    Q3 2024

    • Case management workflows
    • In-app SAR filing forms
    • Cryptographically signed audit log

  3. Phase 3 — ML anomaly detection

    In progress

    Q1 2025

    • Train baseline anomaly model
    • Hybrid scoring (rules + ML)
    • Explainability layer for analysts

  4. Phase 4 — Multi-tenant + reporting

    Planned

    Q3 2025

    • Multi-tenant data isolation
    • Self-serve audit export center
    • Regulator-facing read-only API

Tech Stack

Built with.

ReactNext.jsNode.jsMERN StackTypeScriptPython

Engineering Challenges

Hard problems worth solving.

  • Exposing rule authoring to non-engineers without sacrificing safety
  • Backwards-compatible migration of existing rules from the legacy system
  • Hot-path performance budget under regulatory load

Outcomes

The numbers that matter.

<200ms

P99 scoring latency at 10K tx/min sustained

42%

Reduction in median triage time vs. legacy

0

Compliance gaps in pilot audit window

3 banks

Live in pilot, two more onboarding

Previous case studyZiggasa — Jobs & Education PortalNext case studyStrikeo — E-commerce Management System