Work · 004 · Product + Applied AI
Finance AI
An institutional-grade investment terminal for a private portfolio — and an analyst that has actually read it.
01 The problem
Retail apps answer the wrong question.
A consumer brokerage app is built to tell you whether today was green. A serious holder wants exposure, concentration, correlation to a rate move, and what changed since the last time they looked — the questions a desk asks, answered against their own positions.
And they want to ask them in words, without exporting a CSV into a chat window and hoping.
02 Build
Density over decoration.
Black and amber, tabular figures, no wasted chrome — the visual language of a terminal, because the user is reading numbers in columns, not browsing.
The analyst is the part most people get wrong. It is scoped to the caller's data by the same row-level security that governs the rest of the app, so it can only reason about positions its user is entitled to see. An assistant that can read every portfolio is not a feature; it is an incident.
- Front end
- React, ViteAuth-gated, black/amber terminal UI
- Data
- Supabase — Auth, Postgres, RLS, Edge Functions
- Feeds
- Finnhub, FREDLive market data and macro series
- Research
- TradingView workspace, portfolio news feed
- Analyst
- Claude, portfolio-awareScoped by the caller's own RLS policies