Product concept · Working prototype

Where should your
next dollar go?

Canada gives every saver three tax-advantaged accounts — TFSA, RRSP, FHSA — and no help choosing between them. Room is a two-part product concept that answers the question, then removes the friction that would stop people from asking it.

The problem

The right split across TFSA, RRSP, and FHSA depends on marginal tax rate, home-buying intent, employer matching, and CRA contribution room. Getting it wrong costs real refund dollars every year — over-contributing triggers a 1%/month CRA penalty. Yet most people guess, default to whichever account they opened first, or leave the money in chequing. Canadian fintechs execute trades brilliantly; almost none help with this decision.

What I built

Product 1

Contribution Optimizer

Enter income, monthly savings, province, and home plans — get an explainable allocation waterfall built on 2026 CRA limits and per-province combined marginal rates. Every step carries its "why": employer match first, FHSA if home-buying, RRSP vs TFSA decided by marginal rate. Refund estimate and 10-year projection included.

Open the Optimizer →
Product 2 · AI

NOA Scanner

The optimizer's biggest friction is its first input — almost nobody knows their contribution room. The scanner simulates an LLM document pipeline reading a CRA Notice of Assessment: field-by-field extraction with confidence scores, human-in-the-loop review, then one-tap import into the optimizer.

Open the Scanner →

Product decisions worth noting

1

Explainability is the product. In a regulated space, a black-box "score" is a liability. Every recommendation shows its reasoning, every assumption is user-overridable — which also keeps the tool on the guidance side of the advice regulatory line.

2

The AI reads only what's actually on the document. TFSA room doesn't appear on a Notice of Assessment — so the scanner doesn't pretend to extract it. It's flagged as a default with CRA My Account linkage as the v2 source of truth. AI features earn trust by being honest about their limits.

3

Human in the loop where errors are expensive. A mis-read RRSP limit risks CRA penalties, so extraction shows per-field confidence and requires explicit confirmation before any value is used.

4

The output is an action, not a report. The plan ends in a deposit instruction — the metric that matters to any fintech is funded accounts, not calculator sessions.