> Markdown version of [/jobs/ext/4095318-agentic-ai-developer](https://www.wearedevelopers.com/jobs/ext/4095318-agentic-ai-developer). 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). --- # Agentic AI Developer - **Company:** Blue Electrics - **Location:** Neath, UK - **Salary:** £50,000.0 - **Contract:** Permanent contract - **Skills:** Application Programming Interfaces (APIs), Artificial Intelligence, Business Software, Business Systems, Cloud Computing, Nvidia CUDA, Databases, Data Integration, Software Debugging, Linux, Programming Tools, Issue Tracking Systems, Integrated Development Environments, JSON, Python (Programming Language), Machine Learning, Software Engineering, TypeScript, Software Organization, Retrieval-Augmented Generation, Large Language Models, Claude Code, Prompt Engineering, Qwen, Parallel Computation, AI Coding Agents, Agentic-AI, Git, Build Management, Ollama, Machine Learning Operations, TensorRT, Hardware Infrastructure, OpenClaw (Open-source Robotic Gripper), Invoking Functions, Restful APIs, VLLM, Model Inference, OpenAI Codex, llama.cpp, Webhooks, Deepseek (AI Talent Sourcing Platform), Docker - **Published:** October 5, 2026 - **Apply:** https://www.indeed.com/viewjob?jk=870bff9c46e6a029 ## About the Role We're also very interested in candidates who have experience working with open-weight models and local LLM inference. That could include running and evaluating models from families such as Qwen, Llama, Mistral, DeepSeek or Gemma, and using inference stacks such as vLLM, llama.cpp, Ollama, MLX, TensorRT-LLM or similar. An understanding of the trade-offs between frontier hosted models and locally deployed open-weight models - including latency, privacy, cost, hardware requirements, quantisation and model capability - would be particularly valuable. You should be comfortable with: * Python and/or TypeScript * LLM APIs and tool/function calling * Agent architecture and orchestration * REST APIs and webhooks * Structured outputs and JSON schemas * Open-weight LLMs and local inference * Model serving and inference infrastructure * Prompt and context engineering * Databases and data integrations * Git and modern software development practices * Debugging non-deterministic AI systems Experience with GPU infrastructure, CUDA, quantisation, tensor parallelism, MCP, browser automation, RAG, vector databases, Docker, Linux or cloud infrastructure would also be useful. What matters most We're less interested in whether you have the perfect CV and more interested in whether you build things. If you've spent your evenings experimenting with agents, getting Claude Code or Codex to build software, connecting models to tools, running OpenClaw or Hermes, deploying open-weight models locally, testing inference performance, or trying to make AI actually complete useful tasks - we want to hear from you. We're looking for someone who is curious, technically capable and excited by how quickly this field is moving. ## Description We build AI agents that can actually do work - systems capable of reasoning through tasks, using tools, interacting with software, working across business systems, and completing multi-step processes with minimal human intervention. You'll be working on real-world agentic deployments for businesses, turning manual processes into intelligent, automated workflows. What you'll be working on You'll design, build and deploy AI agents that can: * Reason through complex, multi-step tasks * Use tools, APIs and external systems * Navigate websites and business software * Read and act on emails, documents and structured data * Work with CRMs, ticketing systems, databases and internal platforms * Coordinate multiple agents or specialist sub-agents * Run persistent workflows rather than one-off prompts * Handle failures, retries, approvals and human-in-the-loop processes * Operate reliably in real business environments A big part of the role will also involve figuring out what should be automated, how the agent should approach the task, and how to make it reliable enough for production use. The kind of person we're looking for We're particularly interested in people already experimenting with modern agentic development tools such as: Claude Code, OpenAI Codex, OpenClaw, Hermes, or similar agent frameworks and coding agents. You don't necessarily need experience with every tool on that list, but you should be comfortable working in a world where AI models are not just something you call through an API - they're increasingly part of the development environment itself.