The Lead Applied Scientist will work with large language models, agentic AI, Python, pandas, NumPy, scikit-learn, Databricks, MLflow, and Spark. The role is the senior scientific authority on AXIS Capital’s Data Science and AI Delivery team, building AI/ML solutions that turn data into information for specialty insurance and reinsurance decisions.

Microsoft★4.2 (9k)· London

You will work with Large Language Models (LLMs), machine learning, deep learning, reasoning systems, retrieval-augmented generation (RAG), and reinforcement learning to develop agentic workflows and real-time orchestration platforms. You will join a team of scientists, engineers, and product managers building LLM-based capabilities for Microsoft 365 Copilot, an enterprise productivity product used by millions.

You will work with Java, C++, Python, generative AI, and large models to develop and deploy customized machine learning models at scale. The Prime Video Observability team delivers AI-powered log management and tracing solutions to help developers detect and diagnose issues.

Amazon★3.5 (56k)· London

Develop generative AI and large models using Java, C++, Python, and Unix/Linux for Prime Video’s observability team, which combines log management, tracing, and AI-powered analytics to detect and diagnose issues at unprecedented speed.

Design, develop, and evaluate machine learning models using Python, PyTorch, TensorFlow, statistical modelling, deep learning, and multimodal AI for the ASOS AI Demand Forecasting platform. This platform provides demand forecasts to power replenishment, AI-powered pricing, and supply chain optimisation. You will also mentor other scientists and provide technical leadership on complex modelling projects.

Work with PyTorch or JAX and Python, building tools for data generation, training, evaluation, and deployment on GPU clusters. As an Applied Scientist on Mistral AI's research team, you'll drive pre

You'll work with NLP, LLMs, PyTorch, Hugging Face, and Python on problems in semantic chunking, information extraction, knowledge graph construction, and retrieval optimisation. You'll join an AI research organisation building the document understanding layer that powers next-generation AI search and reasoning systems.

You will work with Python/C++, Ray, Spark, Flink, Kafka/Beam, and ML frameworks for search/recsys and frontier AI (LLMs, RL, multi-agent orchestration). Your role involves building enterprise superintelligence systems—end-to-end architectures that take ownership of business outcomes—as a founding scientist on a small, senior team.

You'll work with Python, PyTorch, AWS, Ray, PyTorch Lightning, and Weights & Biases to design and train machine learning models for Entrust's digital identity verification platform (formerly Onfido), tackling deepfake detection, bias mitigation, document understanding, and anomaly detection. The Applied Science team has approximately twenty machine learning scientists working with product engineering to deploy models for identity verification and facial biometrics at global scale.

Microsoft★4.2 (9k)· London

You will work with Large Language Models (LLMs), machine learning, deep learning, reasoning systems, retrieval-augmented generation (RAG), tool orchestration, reinforcement learning, and telemetry analysis. The team develops advanced LLM-based capabilities and AI-powered experiences for Microsoft 365 Copilot, an enterprise productivity flagship. You collaborate with a team of scientists, engineers, and product managers.

You will work with large language models (LLMs), MCP tools, agent/multi-agent systems, evaluation frameworks, A/B testing, causal inference, NLP, and ML. You will develop and optimize LLM-powered conversational experiences for the Rufus shopping assistant, helping customers find products, compare options, and get recommendations. The London-based Rufus Features Science team includes ~150 engineers, designers, and product managers.

Microsoft★4.2 (9k)· London

Works with Python, PyTorch and retrieval technologies including ColBERT, SPLADE, dense/sparse hybrid, RAG and LLM-integrated architectures. Builds state-of-the-art retrieval systems for Microsoft 365 Copilot Search, Chat and Agent experiences, serving millions of enterprise users daily.

You'll work with PyTorch or JAX, Python, large GPU clusters, deep learning agents, and retrieval-augmented generation pipelines. As part of Mistral AI's research team, you'll run pre/post-training, build datasets and tools, and deploy text, image, and speech models for enterprise clients. No team size is stated.

Amazon★3.5 (56k)· London

You'll use Java, C++, or Python and AI coding assistants to build and ship real-time machine learning models for ranking, valuing, and bidding ads on Amazon's Demand-Side Platform. The team aims to make ads more relevant for shoppers while balancing advertiser and Amazon objectives, working across Edinburgh and London within a larger UK/US group.

The role uses Python, ML, statistics, and LLMs to protect Prolific's platform from fraud by identifying expert participants. As the first data person on the Integrity team, you'll shape the role, collaborate across teams, apply risk modelling or classical ML, turn ambiguous problems into experiments and metrics, and iterate on prototypes to build production solutions.

You will work with Python, PyTorch, TensorFlow and distributed data processing to develop and deploy machine learning models. The AI Demand Forecasting team builds scalable forecasting capabilities supporting replenishment, pricing and supply chain optimisation. No team size is stated.

You will work with Python, classical ML, statistics, LLMs, and agentic techniques (RAG, LLM-as-judge, multi-agent workflows). The role designs and prototypes AI/ML methods to improve human data quality, judgement aggregation, and AI evaluation workflows for Prolific's human data platform.

You'll design and build production models for campaign optimisation using Scala services, SDKs, and data pipelines. The team builds Permutive's data collaboration platform, which uses edge technology and AI to process data securely for privacy-preserving advertising.

Amazon★3.5 (56k)· London

As an Applied Science Intern, you will work with machine learning, deep learning, generative AI, large language models, speech technology, robotics, computer vision, optimization, operations research, quantum computing, and formal methods, using Java, C++, or Python. You will design and develop end-to-end systems, working with Amazon scientists to create algorithms and models that are deployed into production.

Amazon★3.5 (56k)· London

You will work with Java, C++, Python, neural deep learning, machine learning, TensorFlow, PyTorch, Spark, and formal methods (model checking, interactive theorem proving, property-based testing). The Agentic Automated Reasoning Group builds next-generation software verification tools combining AI and cloud computing to analyze code—from Infrastructure as Code to assembly—for provable security, availability, and correctness across AWS services.

Amazon★3.5 (56k)· London

Develops personalised recommendations using Java, C++, Python, neural deep learning, and tools like TensorFlow, Spark, and MLLib. The Prime Video Insights team builds observability features that reduce customer-impact time, consuming petabytes of data daily for internal builders.

Amazon★3.5 (56k)· London

Build machine learning and optimization models for Amazon’s Middle Mile freight marketplace using Java, C++, Python, MxNet, TensorFlow, AWS, and causal inference. The team develops forecasting, dynamic pricing, and network optimization systems to power Amazon’s external freight transportation network.

Amazon★3.5 (56k)· London

The role uses Python, Java, C++, machine learning (random forests, gradient boosting, LASSO, logistic regression) and optimization (integer programming, dynamic programming, stochastic optimization). The team builds algorithms for Amazon's freight marketplace to optimize dynamic pricing and network planning for external shippers and carriers.

Thomson Reuters★3.9 (4.3k)· London

You'll work with PyTorch, JAX, HuggingFace Transformers, vLLM, LLM APIs, Ray, Spark, and experiment tracking tools like Weights & Biases and MLflow on Thomson Reuters' Foundational Research division, which develops advanced algorithms and training techniques for Large Language Models including LLM training, post-training techniques for reasoning and workflows, data-centric ML, and evaluation. You'll design and implement scalable systems for training, fine-tuning, and evaluating LLMs while translating cutting-edge research into production implementations.