Building a Fintech Platform Solo: 185 Tables, 69 APIs, 7 Systems

Most engineers work on one service at a time. I built an entire ecosystem.

The Nexural platform started as a simple idea: a dashboard for my trading community. It became a full fintech platform with 185 database tables, 69 API endpoints, Stripe billing, an AI-powered Discord bot, a research engine, a newsletter studio, and a real-time alert system.

I designed and built all of it. Here's what I learned.

The Scope

Seven interconnected systems:

Trading Dashboard — real-time market data, charts, portfolio tracking
Discord AI Engine — 30+ commands, GPT-4o integration, auto-moderation
Research Engine — 71+ metrics, strategy analysis, CSV import
Alert System — NinjaTrader 8 integration, .NET backend, real-time notifications
Newsletter Studio — automated content generation and distribution
Strategy Tracker — performance monitoring across trading systems
Automation Suite — 61 test suites, CI/CD, quality gates

Database Design at Scale

185 tables sounds intimidating. The key was phased design:

Phase 1 (Core): Users, auth, subscriptions — 20 tables
Phase 2 (Trading): Instruments, positions, signals — 35 tables
Phase 3 (Community): Discord integration, moderation logs — 25 tables
Phase 4 (Analytics): Metrics, reports, telemetry — 30 tables
Phase 5-7: Research, alerts, newsletter — 75 tables

Each phase had its own migration, its own test suite, and its own rollback plan. I never modified more than one domain at a time.

Schema Decisions That Mattered

Normalized where it counts: User → Subscription → Plan is fully normalized. No denormalization shortcuts that would create billing bugs.

Denormalized where speed matters: Trading dashboards query denormalized views. A trader doesn't care about 3NF — they care about sub-50ms load times.

Row-level security everywhere: Supabase RLS policies on every table. A user can never see another user's data, even if the API has a bug.