Backend Engineer (Node.js / TypeScript)

We're building a data-intensive backend platform that ingests large volumes of data, processes and enriches it through a multi-stage pipeline, and uses LLMs to transform it into clean, structured output at scale. The work is well-specified — the data model and the interfaces between stages are already defined — so you can **move fast and work independently**. You'll own significant parts of the pipeline: the orchestration and data flow between stages, the persistence and data-integrity layer, deduplication and record resolution, and the integration of LLM prompts into the system. WHAT YOU'LL BUILD - **The orchestration between pipeline stages** — job routing, queue workers, and the flow that moves a record through each processing stage. - **The data layer on Postgres** (with the `pgvector` extension) — transaction-safe writes, deduplication / record resolution, and the persistence model. - **LLM integration** — wiring prompts into the pipeline with strict JSON / Zod output validation and retry handling. (This is API integration, not ML modeling — but you'll get deep, real-world LLM-pipeline experience.) - **CMS + queues** — Payload CMS collections and BullMQ / Redis job workers. - All of it shipped on **Google Cloud Run**. YOU'RE A FIT IF YOU HAVE: - Solid **Node.js + TypeScript** backend experience — clean, typed, well-tested code. - Strong **SQL / Postgres** — schema design, queries, transactions, and a real instinct for data integrity and concurrency. - Comfort building **REST APIs, async workers / queues**, and integrating third-party services. - The ability to **work independently against a defined spec** and ship quickly — this is an aggressive, focused timeline. - Fluency with Git and a modern dev workflow. BONUS POINTS FOR: - **LLM / GenAI integration** experience (Gemini, OpenAI, Anthropic, etc.) — prompt wiring, structured / JSON output, embeddings. - **pgvector** or other vector-search experience. - **Payload CMS** (or another headless CMS), and **BullMQ / Redis**. - **GCP** (Cloud Run, Cloud SQL) and **Docker**. - **Next.js** familiarity.