Backend AI Engineer

Xist4 IT Limited.; City Of London; Remote; GBP 90,000 - 110,000; Full time

Benefits offered by this job

Fully remote across the UK

Job summary

Xist4 IT Limited is seeking a Backend AI Engineer to design and run high-performance backend services for AI-powered features. You will build production-grade APIs, manage inference pipelines, and ensure low latency across UK-wide remote deployments.

You'll work between models and product clients, using Python and Node.js, with Kubernetes and Docker in the stack. Expect ownership of monitoring, incident response, and scalable architectures in production.

Qualifications

• Backend engineering experience in production environments.
• Experience building high-throughput/low-latency services.
• Proficiency in Python for production backend.
• Experience with LLMs, embeddings or multimodal models.
• Experience debugging distributed systems under load.
• Familiarity with caching, batching or streaming for performance.
• Monitoring and incident response ownership for systems.

Responsibilities

• Build and operate production services for AI-powered features.
• Design clear APIs and service boundaries for mobile and desktop clients.
• Develop inference pipelines and orchestration around LLMs and embeddings.
• Tune throughput and latency with caching, batching, and streaming.
• Own monitoring, logging, alerting and incident response in production.

Job description

Backend AI Engineer | Python, LLM inference, low-latency APIs | Fully remote, UK
£90,000 to £110,000. Permanent.
London. Fully remote across the UK.

Every AI interaction has to pass through a backend that is fast enough, observable enough and cheap enough to survive real use. You will own that layer.

Our client is an early-stage AI product company building applications that get on with everyday tasks before you ask. A prototype exists, launch is ahead, and the company is funded without relying on an upcoming round.

You’ll sit between models and product clients, working closely with ML and frontend engineers. The remit covers inference pipelines, orchestration, service boundaries and the APIs that turn model capability into production behaviour.

This backend gets worked hard. Traffic, latency, batching, caching, streaming and distributed failures are the daily material. When the system slows down or breaks under load, you’ll be one of the people tracing why.

The job

Backend. Build and operate production services for AI-powered features. You’ll design clear APIs and service boundaries that mobile and desktop clients can depend on.
Inference. Build inference pipelines and orchestration around LLMs, embeddings and other model types. You’ll tune throughput and latency across caching, batching and streaming, because a model call does not behave like an ordinary external request.
Operations. Own monitoring, logging, alerting and incident response for the services you build. Production behaviour is the source of truth, and improvements should come from what the system actually does under load.
Engineering. Work mainly in Python and Node.js, with Kubernetes, Docker, SQL and NoSQL in the stack. The job needs backend judgement as much as AI familiarity.

What you’ll bring

Essential:

• Backend services you have built and operated in production.
• High-throughput or low-latency services you have worked on directly.
• Python used for production backend engineering.
• Services you have built around LLMs, embeddings or multimodal models.
• Distributed systems problems you have debugged under load.
• Caching, batching or streaming used to improve service performance.
• Monitoring and incident response for systems you personally owned.
• Existing right to work in the UK.

Useful: Node.js, PyTorch, Kubernetes, Docker, SQL, NoSQL, OpenAI APIs, Anthropic APIs.

Who this suits

You are probably a Backend Engineer, Senior Backend Engineer or AI Infrastructure Engineer who enjoys performance and production behaviour more than feature plumbing. You know what happens after an API meets real traffic.

It won’t suit you if you’d rather call a model like any other API and move on. Here you tune how those calls are batched, cached and streamed, and you are on the incident when they slow down.

We welcome applicants from every background and will support reasonable adjustments.