EXL Service★3.7 (2k)· London

You will work with Python, FastAPI, LLM frameworks (prompt engineering, RAG, agentic systems), and cloud AI platforms (Azure OpenAI/AWS Bedrock/Claude) on Kubernetes. As a Sr. AI Architect & Client Partner at EXL, a data and AI services company, you will design and deliver production-grade Generative and Agentic AI systems for enterprise clients, converting ambiguous mandates into secure, scalable solutions while shaping deals and reusable IP.

You will work with Python, LLM agents, inference/serving, reinforcement learning, multi-agent orchestration, search/retrieval pipelines, and modern AI coding tools. As a DeepMind Research Engineer, you bridge research and scalable systems, collaborating with scientists on machine forecasting to build 0-to-1 infrastructure, optimize LLM architectures and RL pipelines, and engineer evaluation platforms for autonomous reasoning under uncertainty.

This role develops generative AI applications using Java, Python, React or Angular, LangChain, LlamaIndex, Hugging Face, Docker, Kubernetes, and AWS, with tools like GitHub Copilot. You will build scalable, AI-powered solutions for Citi’s global finance operations.

DeepMind★5.0 (2)· London

Work with Python, LLM agents, LLM inference/serving, reinforcement learning for LLMs, multi-agent orchestration, and modern AI coding tools. The Research Engineer builds and scales infrastructure, models, and tools for the Strategic Bets team, transforming research hypotheses into experiments focused on autonomous systems for machine forecasting.

Work with Python, LLM agents, LLM inference/serving, reinforcement learning, multi-agent orchestration, search/retrieval pipelines, simulation environments, and modern AI coding tools (e.g., AGY, Claude Code, Cursor). Design, build, and scale infrastructure for frontier machine-forecasting systems, optimizing LLM agent architectures and RL pipelines to enable autonomous systems that anticipate future events and reason under uncertainty.

You will work with Claude Code, Cursor, Codex and LLMs for prompt engineering, RAG, agents, evaluation and fine-tuning, building full-stack with backend services, APIs, data pipelines and frontend. You will build the client-facing research and sourcing platform for investors and enterprises to find experts, plus internal systems that automate sourcing, screening, matching and managing experts. You will own full slices of product end-to-end from problem definition through deployment.

Citi★3.9 (21k)· London

Build and scale AI-powered financial systems using Java, Spring, React/Angular, Python, FastAPI, Docker, and GitHub Copilot within Citi’s global tech environment. You will design generative AI agents, LLM workflows, and RAG pipelines while working on enterprise microservices, mobile, or front-end development. This role is for an individual contributor on an agile team.

You'll build and fine-tune LLM-powered systems for The Economist's AI Lab, focusing on editorial tone, style transfer, RAG pipelines, and multimodal generation including audio synthesis. Work with Python, LangChain, HuggingFace Transformers, OpenAI/Claude APIs, and evaluation metrics like BLEU and ROUGE to ship zero-to-one products influencing how millions read journalism. You'll be one of three engineers alongside Tech, Design, and Product leads, collaborating directly with journalists on novel AI applications for news.

Accenture★3.9 (29k)· London

You will work with Python, LangChain, LangGraph, AutoGen, FastAPI, REST APIs, vector databases, Git, CI/CD, and cloud platforms (Azure, GCP, or AWS) to build, deploy, and iterate on machine learning and generative AI solutions for clients across industries.

Deliveroo★3.3 (1.5k)· London

You'll work with LLM inference engines (vLLM, SGLang, TensorRT-LLM) and open-weight models (GLM, Qwen, Kimi, DeepSeek) on GPU serving, batch inference, and fine-tuning (SFT/DPO/LoRA), plus build the LLM Gateway, Agent Gateway, evals, guardrails, and cost attribution systems. The GenAI Platform team builds shared infrastructure to bring GenAI products, agents, and automation to production across Deliveroo, DoorDash, and Wolt.

Deliveroo★3.3 (1.5k)· London

You will build production GenAI infrastructure, primarily working with open-weight LLMs/VLMs, GPU serving (vLLM, SGLang, TensorRT-LLM), inference engines, fine-tuning pipelines (SFT/DPO/LoRA), and Python on Kubernetes/AWS/GCP. The GenAI Platform team builds shared infrastructure—including real-time GPU serving, batch inference, the LLM Gateway, and Agent Gateway—to help Deliveroo and DoorDash teams bring GenAI-powered products and automation to production. You join a small, high-leverage team.

You will work with CUDA and Triton kernels, Python, Kubernetes, and AWS to build systems that train and serve large transformer and LLM models. The team creates and scales training pipelines and inference services for Fin’s AI Customer Agent, which resolves customer service queries across channels.

Work with Python, SQL, and ML stacks (PyTorch, TRL) to build evaluation and data pipelines for LLMs and AI coding agents. The team owns how AI agents generate, analyze, and improve Kotlin code across Android, KMP, and server platforms, creating benchmarks and post-training pipelines.

Octopus Energy★3.5 (184)· London

Build and scale Generative AI systems using LLMs, RAG, and agents, working with async Python (asyncio, FastAPI), vector databases (Qdrant, Weaviate, Pinecone, pgvector), orchestration frameworks (LangGraph, LlamaIndex), and AWS. The role is on the Kraken platform team at Octopus Energy, building reusable AI services, knowledge retrieval pipelines, and governance layers to accelerate the clean-energy transition. No team size or scope is stated.

Elliptic★0.0· London

You'll design and build agentic workflows using LangChain or LangGraph, develop backend services in TypeScript and Node.js, and create evaluation frameworks for LLM outputs. Elliptic builds AI-powered compliance tools that help investigators trace blockchain fund flows and detect financial crime. You'll lead the AI team's technical direction, mentor junior engineers, and own significant features from design through production.

You'll design and implement control protocols for Watcher, a monitoring tool for coding agents, using Python, LLM-as-a-judge techniques, fine-tuning, and red-teaming approaches. The team builds empirical safety mechanisms to detect AI agent failure modes in production environments. You'll join a small team with significant ability to shape the tech and product direction.

You'll work with cutting-edge LLMs from OpenAI and Anthropic to build and scale Orbital Copilot, an agentic AI system for commercial real estate legal workflows. You'll own technical direction on multi-agent orchestration, long-context reasoning, retrieval-augmented generation, and evaluation infrastructure, while mentoring the AI engineering team and partnering with the VP of AI and product leadership.

Intercom★3.5 (4)· London

You'll work with GPU optimization (CUDA, Triton), distributed training pipelines for transformers and LLMs, and inference services at scale—building the infrastructure powering Fin, an AI customer service agent resolving millions of queries monthly. The small, highly technical AI Infrastructure team focuses on training and serving custom models like Fin Apex, with responsibilities including kernel tuning, autoscaling, and collaborating with ML scientists to productionize cutting-edge methods. You'll mentor engineers and raise technical standards across Fin's AI platform.

You'll work with machine learning, generative AI, agentic AI systems, prompt engineering, retrieval-augmented generation, and modern frameworks like PyTorch and TensorFlow to deliver large-scale, production-grade AI solutions across healthcare and other sectors. At Kainos, a 150-strong AI and Data Practice, you'll provide technical leadership on complex AI projects, mentor data scientists and AI engineers, and shape commercial AI strategy while engaging C-level stakeholders.

Anthropic★5.0 (1)· London

You'll work across Claude's serving infrastructure—SDKs, APIs, network layers, and accelerators spanning multiple regions and cloud providers—developing service level objectives, designing monitoring systems, and leading incident response. AIRE (AI Reliability Engineering) is Anthropic's cross-cutting reliability team that partners with product teams to keep Claude robust and resilient. Anthropic builds reliable, interpretable AI systems, with Claude as its flagship product.

Anthropic★5.0 (1)· London

You'll work with Python, PyTorch, JAX, and async frameworks like Trio to build reinforcement learning infrastructure and train agentic models at Anthropic, focusing on computer use, code generation, and reasoning capabilities for Claude. The role blends research and engineering, requiring you to architect distributed training systems, design novel RL environments, and optimize performance across GPU clusters while collaborating with alignment and applied teams.