Building an investment product from zero

A B2C investment product — algorithmic crypto portfolios, rebalanced on a schedule, under a CVM securities-consultancy registration — and no engineering team to build it. Investing is a domain where the boring operations are the product: money goes in, money comes out, holdings drift from their targets and have to be pulled back, and every one of those movements has to reconcile to the cent or the company has a problem that is not an engineering problem any more.

The second problem arrived once the first was solved. The product was useful to other companies who wanted investing inside their own apps, and that is a different shape of system than a consumer app with a single front door.

I founded and led the engineering team, and built the core flows end to end: contributions, redemptions, deposits, withdrawals, rebalancing, consolidations. Those are the whole surface of a retail investment product, and they are each other's edge cases — a rebalance is a redemption and a contribution that have to agree.

Serverless on AWS: Lambda behind the Serverless Framework, Node.js and TypeScript, MongoDB, Redis, Cognito for identity. Infrastructure as code from the start, because a small team that deploys by hand stops deploying.

Two decisions worth naming. Kafka came out and AWS SNS/SQS went in — Kafka was infrastructure we were paying for and operating without needing what it is good at, and SNS/SQS integrates with Lambda without a broker to keep alive. And observability went in early: Datadog for logs and metrics, CloudWatch alerts wired to things that actually warranted waking someone.

On mobile, the existing React Native app kept shipping for a year while a Flutter rebuild came up beside it — not a rewrite-and-pray, a running product and its replacement in parallel.

The product shipped and ran, with the full lifecycle in place rather than a subset with the awkward parts deferred.

The B2B integrations opened a revenue stream that had not existed — the same investment capability, sold into exchanges, advisors and family offices instead of only to consumers.

Dropping Kafka reduced both infrastructure cost and the number of things that could break at 3am. Infrastructure as code and CI/CD through GitHub Actions took deployment from an event to a non-event, which is what lets a founding team keep moving.

Node.js, TypeScript, AWS Lambda, Serverless Framework, MongoDB, Redis, Cognito, SNS/SQS (from Kafka), Datadog, CloudWatch, GitHub Actions. Flutter and React Native on mobile.