Senior AI Engineer

Team: Data

Location: London (GB)

Commitment: Full-time

Workplace Type: hybrid

About the role

Octopus was founded with a mission to use technology to accelerate us towards a low-carbon future. That’s why we created Kraken - our own technology platform from scratch which now serves over 70 Million households and is a core reason why Octopus is the number one energy supplier in the UK. We do it by hiring clever, curious, and self-driven people, enabling them with modern tools and infrastructure and giving them lots of autonomy.

We have been using GenAI in live, customer-facing environments since 2022, including one system that creates tens of thousands of high-quality emails for our Energy Specialists, combining our deep knowledge of the energy industry and the Octo communication style with customer-specific data from Kraken.

We are now looking for a Senior AI Engineer. You will build and scale the systems that allow Octopus teams to use Generative AI models (LLMs, RAG, agents). You will be hands-on, working with a cross-functional team to build out flagship AI projects, and the platforms that enable others to succeed with Generative AI.

You’ll work on developing solutions that genuinely move us closer to Net Zero in a company passionate about building great technology to change the way customers use energy. You’ll have wide open problems to solve, so you’ll need to be comfortable with ambiguity, figuring out an approach and validating it fast.

What you’ll do

Design and Develop AI Platform Services: Build reusable, scalable services that expose GenAI models, knowledge retrieval pipelines, and agent workflows to application teams.

Knowledge Base Development: Build and maintain knowledge retrieval systems. Process heterogeneous documents like pdfs and build embedding pipelines that extract, chunk and embed into embedding models.

AI Readiness: Process content from unstructured documents and make it AI ready: Optimize their structure, clear up text ambiguity, and embed rich machine-readable metadata.

AI Ops, evals and observability: Set up framework for monitoring and evaluating AI output quality (relevance, accuracy, safety, drift, cost) and platform observability (latency, cost, usage).

Centre of Excellence: Act as a centre of excellence for the whole business in the technical side of AI and LLMs usage, setting best practices and accelerating adoption.

Governance and Guardrails: Create governance layers for implementing PII redaction, prompt filtering etc.

What you’ll need

Deep GenAI & RAG Expertise: 2+ years of hands-on experience building and productionizing LLM applications, with deep technical knowledge of vector databases (e.g., Qdrant, Weavite, Pinecone, pgvector) and orchestration layers (e.g., LangGraph, LlamaIndex). Comfortable working with heterogeneous documents like pdfs. Experience working with re-ranking models.

Production-Grade Asynchronous Python: Strong software engineering fundamentals with deep experience in asynchronous programming (asyncio, FastAPI) to handle highly concurrent, I/O-bound LLM and database calls. 

Advanced Context Engineering: Proven experience solving production RAG challenges such as context window management, metadata filtering and semantic routing. Experience leveraging coding harnesses like Claude code.

Modern Software Practices: Mastery of Git, automated testing and CI/CD pipelines. Experience working with cloud platforms like AWS.

Comfort with Ambiguity: Ability to rapidly prototype an approach, validate it with metrics, and pivot fast in a fast-moving environment.

It would be great if you had

GraphRAG Experience: Familiarity with Knowledge Graphs (e.g., Neo4j) and ontologies to enhance standard vector RAG with highly structured, interconnected corporate data.

Local/Open-Weight Model Deployment: Experience deploying and fine-tuning open-weight models (like Llama 3 or Mistral) via frameworks like vLLM or Ollama to optimize token costs and privacy.

Prompt Caching & Cost Optimisation: Practical experience implementing semantic caching layers to drastically reduce LLM API billing and response latency.