Selected work

Case studies

Hands-on production engineering and the architecture decisions behind it.

Dynamic funnels Multi-brand platform

Fintech platform for SMB financing

Dynamic onboarding funnels

My role: full stack implementation · Next.js, React, Zod, NestJS, TypeORM, PostgreSQL

~20%: reduction in form drop-off

Flow configurationSteps and branching depend on the business profile and the qualification answers.

Rendered steps

Next.js and React render the step the applicant is on.

Dynamic Zod schemasEach step is validated on the client against a schema built for it.

Partial draftsProgress is saved step by step, so an applicant can leave and resume.

PersistenceNestJS and TypeORM store the application entities in PostgreSQL.

How an application moves through the funnel

Problem

SMBs apply for financing through multi-step onboarding and loan-qualification forms. The forms are long, the steps differ by business, and applicants were dropping off before finishing.

What I built

Multi-step funnels in Next.js and React, with client-side validation from dynamic Zod schemas. On the back end, NestJS and TypeORM persist the application entities in PostgreSQL.

The hard part

Partial drafts. Applicants needed to save their progress and come back later, but a half-finished form fails strict schema validation. The draft persistence had to accept incomplete steps across the whole multi-step form without raising those errors.

Result

Form drop-off on the qualification funnels fell by about 20%.

Multi-brand fintech company

Multi-brand platform

My role: back-end services and tenant isolation · NestJS, TypeORM, PostgreSQL, background queues

~60–70%: less recurring manual operational and sync work across partner brands
7 to 10: partner brands on one platform

Brand channelsEach brand connects several channels that send data in.

Channel servicesNestJS services receive each channel’s data and process it.

Background queuesEvents are processed asynchronously. Each payload carries its brand, so third-party calls route correctly.

Tenant scoping

NestJS guards and TypeORM repositories filter every query by brand ID.

Dashboards and accessDashboards, user management and permissions for each brand.

How a brand’s data moves through the platform

Problem

The company ran 7 to 10 brands, each with its own channels producing data. They needed one platform to work in, but every brand had to see and act on its own information only.

What I built

Back-end services in NestJS with TypeORM on PostgreSQL. Events are processed asynchronously through background queues, and each event payload carries its brand context so calls to third-party services route to the right brand.

The hard part

Tenant isolation. I implemented it at the application level: NestJS guards establish the brand for each request, and TypeORM repositories scope every query by brand ID. We used this rather than database-level Row-Level Security or a database per tenant.

Result

Recurring manual operational and sync work across the partner brands dropped by roughly 60 to 70%, with strict multi-tenant data isolation across services in production.

Contact

Let’s talk.

I’m looking for a Full Stack Engineer role. I’m based in Israel and open to hybrid or remote. Email is the fastest way to reach me.

raday146@gmail.com