Backend Engineer (AWS-based SaaS Platform, LLM-powered)*
Project: 4- to 6-week build sprint
Location: Remote (2–3 h overlap with UK time)
Fee: Fixed-price USD $4,000-$6,000
About Us
We’re a boutique consultancy spinning up an AI-enabled SaaS product. Our AWS backbone—multi-AZ VPC, ECS Fargate, ALB, Terraform IaC, logging/monitoring, IAM hardening, etc. is already in place. Your brief is to layer business logic, API endpoints and a small admin UI on top of this platform, integrating an LLM-driven scoring pipeline plus a few external APIs.
What You’ll Tackle
*Backend feature work*
• Design and ship REST or GraphQL endpoints.
• Build multi-tenant data models and role-based workflows.
• Trigger emails and webhooks for status changes.
*AI / Retrieval-Augmented Generation*
• Wire existing prompts to an LLM API (OpenAI-compatible today, private model later).
• Store embeddings in a vector store and perform RAG look-ups.
• Surface “explanation” JSON for every model decision.
*Cloud hand-offs*
• Deploy new services into the existing Terraform / ECS stack—extend, don’t rewrite.
• Add CloudWatch alarms, log routes and parameter-store secrets where needed.
*Quality & Ops*
• Write unit and integration tests and set up a lightweight CI pipeline.
• Provide concise hand-over docs and 30 days of post-launch bug-fix cover.
Must-Have Skills
• Pro in *Python (FastAPI)* or *Node.js (TypeScript / Nest)*—ready to justify your pick.
• Experience extending an *AWS stack (Terraform, ECS Fargate, ALB, Secrets Manager, KMS)*.
• Hands-on with *LLM APIs* and at least one *vector database* (Pinecone, Weaviate, OpenSearch, etc.).
• Multi-tenant data design with GDPR awareness.
• CI/CD and automated-testing mindset.
Nice-to-Haves
• *Solution-architect background*—ability to map future services and scaling paths.
• Prior LLM fine-tuning or custom-model training.
• Chrome-extension data capture.
• Transactional-email tooling (SES, Postmark, Mailgun).
Following this sprint project there will be an additional 2-3 sprint projects of similar size to follow.