> Markdown version of [/jobs/ext/4017093-ai-lead-developer](https://www.wearedevelopers.com/jobs/ext/4017093-ai-lead-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). --- # AI Lead Developer - **Company:** RWS Group - **Location:** London, UK - **Experience:** Expert - **Contract:** Permanent contract - **Skills:** Artificial Intelligence, Amazon Web Services, Computer Vision, Software Quality, Code Review, Python (Programming Language), Language Modeling, Natural Language Processing, Regression Testing, Management of Software Versions, Visual Systems, Pytorch, Retrieval-Augmented Generation, Large Language Models, Prompt Engineering, Technical Debt, Agentic-AI, Build Management, Kubernetes, HuggingFace, Machine Learning Operations, Software Version Control - **Published:** October 30, 2026 - **Apply:** https://www.indeed.com/viewjob?jk=577f810616a3f99b ## About the Role * Commercial experience as an AI, ML or computer-vision engineer, including building, or helping to build, AI systems that are running in production. * A track record of technical ownership: making architectural calls on AI systems. * Experience leading, mentoring or technically guiding other engineers, formally or informally, and a clear view of how a strong AI team should work. * Experience establishing or improving team practice, such as evaluation standards, release processes, shared tooling or ways of working. * Strong Python, with deep working knowledge of PyTorch and the Hugging Face ecosystem. * Experience with automatic or model-assisted labelling of visual data, and with dataset quality control. * A solid grasp of MLOps: experiment tracking, model and dataset versioning, reproducible training and containerised deployment. * Strong communication skills, including explaining estimates, risks, trade-offs and model limitations. * Experience building with LLM APIs, including prompt engineering, structured output, RAG and provider trade-offs. * Practical computer-vision experience: detection, classification, and image similarity, including fine-tuning on custom datasets. Nice to Have * Experience with zero-shot and vision-language models (e.g., OWLv2, Florence-2, CLIP-style embeddings). * Experience building AI agents, tool use, or computer-use and browser automation. * Experience with ClearML, Label Studio, PyTorch Lightning, or LangChain. * Experience running GPU inference on Kubernetes, AWS (including Bedrock), or GCP. * Hands-on experience designing evaluations for vision and/or LLM systems, including baselines, metrics, test sets, and comparisons between versions. ## Description You'll be accountable for our AI systems end to end, lead the AI team, and shape how that team designs, evaluates, ships, and runs AI. You'll stay hands-on, building and designing AI solutions day to day, while setting the direction and standards others work to. This role would suit a senior engineer ready to step into leadership, with a track record of building production-grade applications across a range of AI/ML disciplines such as computer vision, NLP and agentic systems., System ownership * Own the AI estate end-to-end, across vision models, ML pipelines, LLM services, and AI agents. You are accountable for how it's designed, how it performs in production, and how it evolves. * Set and maintain the technical direction and architecture for AI. Decide when to build, fine-tune, use a foundation model, or call a provider, and document those decisions so they're understood and applied consistently. * Own production quality, cost, and reliability for AI: monitor drift, accuracy, latency, and spend, and act on what the data shows. * Identify and drive down technical debt across the AI estate. Turn one-off scripts, notebooks and proofs of concept into shared, maintainable systems. Leading the team and shaping how it works * Lead the AI team day to day: set priorities, unblock work and make sure the team delivers what it commits to. * Define how the team works, from problem framing and data collection through labelling, evaluation, release and monitoring, then write it down and make it the default. * Make sure every model and AI feature in production is versioned, observable, and traceable, so we can always say what produced a result, when, and how well it's performing. * Establish shared tooling, patterns and standards for the team: dataset and model versioning, evaluation harnesses, code quality and reproducibility. * Mentor and develop engineers and data scientists through code review, pairing and knowledge sharing, and help each person grow. * Build onboarding material, documentation, and reusable examples so new joiners are productive quickly. Hands-on design and delivery * Design and build AI solutions yourself, including computer-vision models, LLM-powered features, retrieval-augmented generation and agentic workflows, and lead the team's most technically demanding work. * Own our approach to evaluating AI and vision systems: benchmark sets, metrics, regression testing between versions, and human-in-the-loop review. * Lead our approach to automatic and model-assisted labelling of images, using foundation and zero-shot models to speed up dataset creation without losing quality. Commercial and stakeholder leadership * Be the point of contact for AI across product, engineering and client-facing teams. Turn client problems into AI work with clear success measures. * Understand the commercial impact of AI decisions, including provider and compute cost, accuracy trade-offs and the review workload they create, and make the case for investment where it's needed. * Give realistic estimates for research and delivery, flag risks early, and know when to stop a proof of concept. * Explain AI capabilities, limitations, and recommendations clearly to technical and non-technical audiences, including clients and senior leadership. * Judge which AI/ML approach best fits each situation, and pull together the resources we already have to build effective solutions, for example, combining computer vision and NLP into vision-language model (VLM) solutions for client demo labelling.