> Markdown version of [/jobs/ext/2972687-senior-ai-software-engineer](https://www.wearedevelopers.com/jobs/ext/2972687-senior-ai-software-engineer). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md). --- # Senior AI Software Engineer - **Company:** Centrica - **Location:** Windsor, UK - **Experience:** Expert - **Contract:** Permanent contract - **Skills:** Artificial Intelligence, Amazon Web Services, Computer Literacy, Continuous Integration, Software Debugging, Memory Management, Interoperability, Python (Programming Language), Routing, Productivity Software, Software Engineering, Data Streaming, Core Voice Platform, Amazon Connect, Datadog, Large Language Models, Model Validation, Caching, Build Management, AI Platforms, Information Technology, Production Code, Terraform, Automation Anywhere, Serverless Computing, Human in the Loop - **Published:** September 18, 2026 - **Apply:** https://www.totaljobs.com/job/senior-ai-software-engineer/centrica-chp-job107994577 ## About the Role * Substantial software engineering experience, with strong Python. You write production code, not just notebooks. * Demonstrable delivery of LLM or agentic systems into production for real users, and the scar tissue that comes with it. We will ask what broke and what you changed. * Hands on experience with tool and function calling, multi step agent orchestration, and managing conversational state across turns. * Production experience with retrieval augmented generation, including retrieval quality evaluation and answer grounding, not just wiring up a vector store. * A rigorous approach to evaluation. You have built or owned an eval harness, defined the metrics that mattered, and used them to make release decisions. * Experience deploying and operating AI services in the cloud using serverless or container based architectures, infrastructure as code such as Terraform, and CI/CD. * Practical application of responsible AI, covering data privacy, guardrails, bias, explainability and human in the loop design, in an environment with real compliance obligations. * Ability to debug systems that are non deterministic by nature, and the patience to reason about behaviour statistically rather than case by case. * Strong collaboration and communication skills. You can explain to a product owner why the agent got something wrong, and to a governance board why the controls are sufficient. * Degree in Computer Science, Engineering or a related discipline, or equivalent practical experience. Desirable: * Amazon Bedrock specifically, including AgentCore Runtime, Bedrock Knowledge Bases and Bedrock Guardrails. If your experience is on another foundation model platform, that transfers. * Voice AI experience, including speech to speech models, real time streaming, barge in and turn taking, or ASR and TTS pipelines. * Any contact centre exposure, including Amazon Connect. A bonus rather than an expectation. * Familiarity with agent interoperability standards such as the Model Context Protocol. * Experience with fine tuning, distillation or model routing where it materially moved cost or latency. * Delivery in a regulated industry, including working with vulnerable customer requirements or comparable obligations. * AWS certification at Associate level or above. * Core Competencies & Technical Skills (AI and emerging technology) * Designs, integrates and operates AI-enabled solutions within enterprise environments, including prompt-driven workflows, retrieval-augmented systems and AI agents. * Applies structured evaluation, testing and monitoring practices to ensure AI outputs are reliable, secure and compliant with organisational guardrails. * Prepares and manages data used in AI workflows and takes responsibility for the responsible lifecycle of AI features from experimentation through deployment and continuous improvement. AI / Digital fluency * Demonstrates the safe and responsible use of AI tools, with clear knowledge of when AI use is appropriate and strong awareness of accuracy, bias and compliance. Bringing the ability to design and reuse prompt templates to support consistent, high-quality workflow outputs, and skilled in using AI to triage, classify and analyse information within Centrica policy guardrails. * Strong ability to recognise higher-risk scenarios and escalate to governance or security as needed. Alongside this, showing proficiency in enterprise AI co-pilots, knowledge assistants and AI-enhanced productivity tools. ## Description This is a role for someone who has taken generative AI past the demo stage. Our problems are the hard ones that show up after launch: grounding answers in the right source, keeping voice latency low enough to feel natural, handing off to a human at the right moment, and proving to a governance board that the system behaves safely at scale. You do not need a contact centre background. If you are a strong engineer who has built real agentic systems on Bedrock or a comparable platform, we will teach you the domain., * Design and build agentic services on Amazon Bedrock and AgentCore Runtime, including multi step orchestration, state and memory management, and graceful handling of failure and ambiguity. * Build and own the tool layer that agents act through, exposing business capability such as account lookup, billing and identity as reliable, well described actions with sensible error semantics. * Build retrieval augmented generation over Bedrock Knowledge Bases, taking ownership of content preparation, chunking, embedding strategy, retrieval quality and answer grounding. * Treat prompts as engineering artefacts. Version them, test them, and understand precisely what changed when behaviour shifts. * Build the evaluation harness. Define what accuracy, containment and safe behaviour mean for each product, assemble regression sets, and make it possible to ship a prompt or model change with evidence rather than optimism. * Implement guardrails and safety controls, including PII handling, out of scope refusal, jailbreak resistance and clean escalation to a human advisor. * Engineer for latency and cost. Streaming, model selection and routing, caching and token budgets all matter, and in voice they matter in real time where a pause of a few hundred milliseconds is felt by the customer. * Instrument agent behaviour end to end so that tool failures, retrieval misses and abandoned conversations are visible in Datadog and QuickSight, not discovered through complaints. * Support live AI services in production, investigating incidents and diagnosing non deterministic behaviour, then closing the loop with permanent fixes. * Make and document build versus buy calls, writing Architecture Decision Records and taking designs through our AI governance board and change control process. * Track how the generative AI landscape is moving and bring the parts that deliver clear benefit into the roadmap, with an honest read on the risk as well as the upside. ## Related Videos - [AI-Enabled Organisations: From Strategy to Practice](https://www.wearedevelopers.com/videos/100171-ai-enabled-organisations-from-strategy-to-practice) - [Debugging in the Dark](https://www.wearedevelopers.com/videos/1658-debugging-in-the-dark) - [Creating a routing app with Google Maps API from scratch](https://www.wearedevelopers.com/videos/831-creating-a-routing-app-with-google-maps-api-from-scratch) - [HTTP headers that make your website go faster](https://www.wearedevelopers.com/videos/1676-http-headers-that-make-your-website-go-faster) - [Reducing LLM Calls with Vector Search Patterns - Raphael De Lio (Redis)](https://www.wearedevelopers.com/videos/1714-reducing-llm-calls-with-vector-search-patterns-raphael-de-lio-redis) - [Agentic employees in world's most downloaded FinTech app](https://www.wearedevelopers.com/videos/100123-agentic-employees-in-world-s-most-downloaded-fintech-app) ## Related Articles - [Coffee with Developers - Maria Apazoglou - Making AI understandable for all in production](https://www.wearedevelopers.com/magazine/475-coffee-with-developers-maria-apazoglou-making-ai-understandable-for-all-in-production) - [From Prototype to Production: Build AI Agents with This Free 4-Course Learning Path](https://www.wearedevelopers.com/magazine/655-from-prototype-to-production-build-ai-agents-with-this-free-4-course-learning-path) - [Navigating the AI Shift](https://www.wearedevelopers.com/magazine/629-navigating-the-ai-shift) - [How to Become an AI Engineer](https://www.wearedevelopers.com/magazine/331-how-to-become-an-ai-engineer) - [Graph and AI Trends 2026: Why Is AI Running but Not Yet Delivering?](https://www.wearedevelopers.com/magazine/680-graph-and-ai-trends-2026-why-is-ai-running-but-not-yet-delivering) - [Got AI ideas but no money? Here are 10 free ways to level up your AI skills with Google Cloud](https://www.wearedevelopers.com/magazine/600-got-ai-ideas-but-no-money-here-are-10-free-ways-to-level-up-your-ai-skills-with-google-cloud)