Back to projects
|
Case study · Fintech · Founder

Tesora

AI-powered personal finance: connect your banks, unify your wealth, get intelligent advice.

Role
Founder · Lead full-stack
Status
Live
FintechNext.jsReactTypeScriptOpenAIPostgreSQL
Thesis

One place to see, understand and move your money.

One app — accounts, debts, investments, and an assistant that understands the full context.

01
The problem

Existing personal-finance products tend to live at two extremes: cold aggregators that only display numbers, or generic coaching apps with no real access to your situation.

Tesora was born to close that gap — a serious aggregator with an AI assistant that actually knows your wealth, your debts, and your investments, and can reason over them in real time.

02
Tech stack

App Router + RSC + Server Actions. The layer holding up the whole frontend and half of the backend.

useOptimistic and transitions so financial mutations feel instant without lying to the user.

Strict mode across the board, input validation with Zod before anything hits the DB.

Supabase + Drizzle. Real Row Level Security — each user only sees their own rows, including from the AI.

GPT with tool-calling over a read-only view of net worth. The model reasons, it never touches credentials.

Subscriptions, billing for the advisor marketplace, and Connect to split payments.

Design system with semantic tokens; zero loose CSS inside components.

Continuous deploy + Edge runtime for low latencies across LATAM and Europe.

Tap or hover any technology.01 / 08
03
Key features
01

Secure account aggregation and unified net worth, with monthly history.

02

Smart expense analysis and automatic categorization.

03

Debt management with Avalanche and Snowball simulations.

04

Marketplace of certified financial advisors, with integrated billing.

05

Conversational AI assistant with contextual access to your data.

06

Savings planning with realistic scenarios and projections.

07

Dynamic Financial Health Score and proactive email alerts.

04
Architecture

I keep things reasonably flat: a Next.js frontend calling server actions and an Edge worker that orchestrates the assistant. The AI layer has read-only access to a view of the user's data — never to banking credentials.

05
Learnings
With Tesora I've learned to see my money in a completely different way: no longer just scattered balances, but how income, expenses, debts and investments relate in real time. And technically, I've internalized that useful AI isn't the biggest model — it's the one with the right context.

Links

Other projects