Full-Stack AI Engineer

Role Description We are seeking a senior Full-Stack AI Engineer to join our platform team and own the end-to-end build of a data ingestion and intelligence layer for enterprise customers in the media vertical. This is a high-impact, high-ownership role at the intersection of AI, data engineering, and product. You will architect and ship pipelines that transform unstructured creative assets and ad performance data into a semantic, vector-based layer — and expose it through AI agents and a polished React frontend. We move fast, ship to production, and hold a high bar for craft and reliability. What You'll Build A dual-mode data ingestion engine handling creative components (CTAs, headers, body copy, images, video, and metadata) and ad performance data (channel-level signals tied to those components) Multi-modal embedding generation and storage — vector plus structured — optimised for retrieval quality and cost AI agent tooling that enables natural language search, comparison, and reasoning across the creative and performance data layer A React frontend that lets users explore the creative library, query performance data in plain English, and surface actionable optimisation insights Audit-logged decisioning and governance infrastructure to meet enterprise-grade requirements Key Responsibilities Design and build durable, idempotent ingestion pipelines for creative and performance data at scale (queues, retries, backpressure, dedup, schema evolution) Generate and manage embeddings for multi-modal creative assets; select and operate the right vector store for the workload Build and maintain retrieval pipelines that serve AI agent tools with accurate, low-latency responses Ship agent-style systems with tool use, state management, and multi-step reasoning workflows Develop and maintain the React frontend for the creative intelligence library and query interface Own the full lifecycle of your systems: design, build, deploy, monitor, and iterate Contribute to stack decisions with clear reasoning grounded in production experience Collaborate closely with product and enterprise partners to translate requirements into reliable, scalable systems Qualifications Strong TypeScript — you use types as a design tool, not a formality Production experience with serverless or edge runtimes (Cloudflare Workers, Vercel, Lambda, Deno Deploy, or equivalent) Demonstrated experience building durable, idempotent ingestion pipelines with queuing, retry logic, backpressure handling, deduplication, and schema evolution Practical, production-level understanding of embeddings, chunking strategies, and retrieval quality tuning At least one agent-style system shipped to production: tool use, stateful multi-step workflows — framework matters less than the experience React fluency with modern patterns and component architecture Comfort operating across two cloud environments; able to reason clearly about when to use edge compute vs. managed data/AI services, and how to bridge them Must have prior remote work experience, be fluent with remote collaboration tools and platforms (such as Slack, Zoom, Google Workspace, Linear, or similar), and have ideally worked with US or UK-based companies. Applications without this experience will not be considered. Preferred Qualifications Experience building or operating RAG systems in production Familiarity with current embedding models and the tradeoffs across dimension, quality, and cost Background in ETL design, observability for data pipelines, or evaluation frameworks for retrieval quality Adtech, performance marketing, or marketing analytics background — understanding what channels, attribution, and creative testing look like in a live production context Opinions on vector databases (Cloudflare Vectorize, Vertex AI Vector Search, Turbopuffer, or similar) backed by hands-on experience Tools & Technology TypeScript (primary language across the stack) Cloudflare Workers, Queues, and Agents SDK (or equivalent edge runtime) GCP — Vertex AI for embeddings and related data/AI services Vector database (to be selected: Cloudflare Vectorize, Vertex AI Vector Search, Turbopuffer, or similar) React with Remix or TanStack Start (TBD) Google Workspace, Slack, Zoom, and standard remote collaboration tooling Please NOTE It is crucial that you complete the application form in full. As part of the application process, you will be required to record a video. If your application is successful, you will receive an email confirming next steps — the video is the first step of the interview process. If you do not record a video, we will not be able to consider you for ANY open roles.