60 jobs
PhysicsX San Francisco, United States US$150,000 – US$190,000 pa
Delivery-focused role embedding physics-aware AI models into customer tools, building reliable ML data pipelines and working with 3D point-cloud and mesh data. Spans data science and software engineering, translating R&D into reusable libraries and products, with regular on-site customer travel.
PhysicsX New York, United States US$150,000 – US$190,000 pa
Builds and deploys machine learning systems for industrial physics challenges, designing reliable data pipelines and working with 3D point-cloud and mesh data. Partners directly with customers on-site to embed AI models into practical tools, translating R&D outputs into reusable libraries and products. Requires Python, cloud, and MLOps experience with frequent travel.
Cohere London
Research scientists and engineers advancing multilingual large language models, designing scalable solutions to improve LLM performance across many languages and skills. The role spans building ML pipelines, publishing research at top-tier venues, and mentoring junior team members, with a strong emphasis on Python and NLP fundamentals.
Cohere London
Hands-on research and engineering role on the Agent Code team, developing code-generating LLMs and autonomous agent systems for enterprise use. Involves training code models, building agent frameworks for inference and sampling, creating SFT trajectories, working on RL algorithms, and designing benchmarks, in collaboration with the pretraining team on frontier compute infrastructure.
ElevenLabs London
Builds and operates the platform every team ships on, running Kubernetes clusters including GPU fleets for training and inference across cloud and on-prem. Sets up compute, networking, IAM and data systems for other teams, and writes Python or Go tooling to make infrastructure self-serve. Works hands-on in a flat team on cluster builds and cloud migrations.
ElevenLabs United Kingdom
Owns the data infrastructure behind frontier AI models, building large-scale pipelines for collecting, filtering and transforming training datasets. Also trains classifiers, quality filters and labeling models, and designs curation strategies like deduplication and quality scoring. Fully remote, with tooling work letting researchers explore and train on massive datasets.
PhysicsX London
Own a research work-stream at the frontier of 3D geometry representation and generation, designing and training generative models over geometry and closing the coupling between geometry and physics. Work against production numerical solvers so model quality is measured in validated physics, collaborating with ML and simulation engineers and customers.
Cohere London
Own end-to-end technical program delivery for UK public sector and defence customers, from initial deployment through production scaling, acting as the bridge between frontier AI capabilities and government requirements. Navigate procurement, security authorisations, data sovereignty and compliance while coordinating Applied ML, Forward Deployed Engineering, Platform and Product teams. Work spans air-gapped, on-premises and highly regulated environments.
Cohere London
Owns the evaluation strategy for an enterprise AI workspace, building evals that capture agentic workflows, tool use and knowledge work. Translates user feedback, production failures and usage logs into continuous measurement, then feeds model failures and capability gaps back to central modelling teams. Sits between product, customer-facing and research teams rather than being pure research or product engineering.
Cohere London
Builds reinforcement learning environments for enterprise AI agents, populating them with hard tasks, plausible data and reward verifiers, then runs agents inside to evaluate and train them. Works across modelling and product to find capability gaps, and partners with external vendors on expert-built environments and quality tooling.
Cohere London
Designs and writes high-performance, scalable software for training frontier models, improving the training setup from an infrastructure and codebase performance standpoint. Builds tooling to speed up training cycles and works on supercompute and data infrastructure, bridging research and production. Works with distributed training stacks like Kubernetes, Slurm and Ray, with one of the highest compute-to-engineer ratios in the world.
Cohere London
Builds and maintains the data pipeline behind Cohere's large language models, running data ablations to assess quality and designing pre-training data mixtures that improve model performance. Combines research and engineering, working with Python, Spark or Beam on web, code and multilingual corpora alongside researchers and engineers.
Cohere London
Research engineering role on the integration team, building and scaling machine learning algorithms and infrastructure for LLM post-training, with a focus on large-scale distributed RL methods. Day to day involves writing high-performance training software, developing research tooling, and collaborating with infrastructure, efficiency, serving and scientific teams to improve SFT, offline preference and RL training cycles.
Cohere London
Forward deployed engineer building and shipping LLM-powered agentic workflows for enterprise customers, from early prototypes to production-grade agents. Works directly with clients in finance, healthcare and telecoms, integrating models with tools, APIs and sensitive data, and contributing features to Cohere's North AI workspace platform.
Cohere London
Designs next-generation evaluation benchmarks and infrastructure to measure LLM progress, including training LLM judges, refining data synthesis pipelines, and improving evaluation efficiency. Works cross-functionally to turn model feedback into trustworthy, repeatable evaluations, and builds scalable tools for probing model performance.
