About the product
Brand licensing is a massive global industry that still runs on introductions and paperwork. At Negosh,
we're rebuilding it around AI. Brands and manufacturers find each other on our marketplace, an in-app
assistant walks them through structuring and closing the deal, and we work with some of the biggest
brand owners in the world, helping them automate and improve their licensing programs. Very little
software has ever reached this market, so almost everything we ship is the first version of itself.
The role
You'll own features end-to-end across a TypeScript codebase on both sides of the wire: a Fastify +
Prisma backend and a Next.js + React frontend. A typical feature runs from a Prisma schema change and migration, through Zod-validated Fastify routes, to React Query hooks and MUI views — and ships with route-level integration tests that run against real Postgres and Redis in containers. If that sounds like your kind of loop, read on.
- Design and ship full-stack features across ~40 backend domain modules and a large component- driven frontend.
- Build and extend LLM-powered features: structured-output prompting, evidence gating, embeddings and vector search (pgvector), token budgeting, and cost tracking.
- Work on high-stakes negotiation flows: deal memo revisions, renewals, e-signature (DocuSign), payments (Stripe, Plaid, Tipalti).
- Extend the multi-tenant white-label architecture — per-site theming, custom-domain SSO
(OIDC), site-scoped data.
- Maintain quality through heavy automated testing: Mocha + Testcontainers integration suites on the backend, Vitest + React Testing Library on the frontend.
Our stack
BACKEND
TypeScript · Node.js · Fastify · Prisma · PostgreSQL + pgvector · Redis · Elasticsearch · Zod · OpenAI SDK · Cloudflare R2 · Docker
FRONTEND
Next.js · React 18 · TypeScript · MUI + Emotion · TanStack Query · Zustand · react-hook-form · next- i18next · Stripe.js
TESTING & DELIVERY
Mocha + Testcontainers · Vitest + RTL · GitHub Actions · TeamCity · Sentry · Sinon
Must have
AI - Augmented Development
This one really matters to us: you use AI tools to augment your development, and you're good at it.
Agentic coding tools (Claude Code, Cursor, Copilot, or similar) are a daily part of your workflow — you
direct them well, review and validate what they produce, and know when not to trust them. We invest a lot in making our codebase legible to both humans and AI agents (convention docs, a linted project wiki, strong test guardrails), and you'll help us keep that going.
- 5+ years of professional full-stack experience, with deep TypeScript on both the server and the client.
- Production Node.js API experience — designing REST endpoints, request/response validation, auth (JWT/OIDC), background workers, and queues.
- Production React experience — component architecture, server-state management (React Query or similar), forms, and responsive UI in a large codebase.
- Strong relational-database skills — schema design, migrations, and query performance in PostgreSQL (we run ~60 models and 400+ migrations).
- Testing discipline — you write integration and unit tests as part of shipping, not after.
- Ownership — you can take a loosely specified feature from a ticket to production, including the edge cases, the tests, and the docs. Since we're fully remote, this kind of self-direction is how we all work.
Nice to have
- Experience building LLM product features: structured outputs, embeddings/RAG, agentic pipelines, prompt evaluation, cost control.
- Fastify, Prisma, MUI, or Next.js specifically.
- Multi-tenant SaaS, payments infrastructure, or e-signature integrations.
- Internationalization, media processing pipelines (ffmpeg, sharp), or search (Elasticsearch, vector search).
How we work
We're a small, senior, fully remote team — high trust, high autonomy, and no one looking over your
shoulder. Every change ships behind CI gates (typecheck, lint, and full test suites on both repos), and
backend integration tests run against disposable Postgres and Redis containers, so tests exercise real
behavior rather than mocks. We document architecture decisions in a source-backed project wiki that
both engineers and AI agents rely on, and we treat that documentation as part of the definition of done. Work from wherever you do your best work.
