Skip to content

AI & LLM

Ellavator AI (Ella)

Full-Stack Engineer on an AI Marketing Platform

Role
Full-stack engineer
Timeline
Ongoing
Team
Product team
Live link
www.ellavator.ai
Multi-tenant
brand-scoped workspaces
1
model-agnostic LLM layer
Live
production platform

The problem

Marketing teams using general-purpose AI tools produce more content but lose what makes each brand distinct. Ellavator needed a platform where each brand's strategy lives as persistent context the AI works from, with strict isolation between brands so an agency can run many clients in one workspace with zero data bleed.

My approach

Multi-tenant architecture

Built on PostgreSQL via Supabase with each brand as an isolated workspace holding its own strategy, offers, proof, and assets. Fine-grained authorization via WorkOS FGA resolves permissions per brand rather than per account.

AI pipeline

Assembled brand-specific context into prompts so output reflects each brand's positioning and voice, with persistent conversation history for full continuity across sessions.

Model provider migration

Led the production migration from a direct OpenAI integration to OpenRouter, abstracting the provider behind a common interface — removing single-vendor dependency and opening model selection on cost and capability.

Backend & integrations

Nest.js/TypeScript API services, Stripe for billing and entitlements, HubSpot for CRM sync, API Deck for third-party integrations, Sentry for observability.

Frontend

Next.js/TypeScript with TanStack Query for server state — caching, background refetching, and optimistic updates, which matters when LLM responses stream and conversation state must stay consistent.

System shape

  1. Frontend

    Next.js, React, TypeScript, TanStack Query

  2. Backend

    Nest.js, Node.js

  3. Data

    PostgreSQL, Supabase

  4. Infra

    Sentry

  5. AI

    OpenAI API, OpenRouter, LLM Integration

Stack

Frontend

  • Next.js
  • React
  • TypeScript
  • TanStack Query

Backend

  • Nest.js
  • Node.js

Data

  • PostgreSQL
  • Supabase

Infra

  • Sentry

AI

  • OpenAI API
  • OpenRouter
  • LLM Integration

Result

A live production platform used by agencies, fractional CMOs, and in-house marketing teams, running brand-scoped AI generation across many isolated client workspaces, with subscription billing, CRM sync, and a model-agnostic LLM layer.

Next projectHeroGeneration