Research Engineer
2 days ago
City of London
Research Engineer – Foundation Models | London (On-site) We’re partnering with a stealth-mode deep tech AI lab based in London that’s building foundation models from the ground up. This isn’t fine-tuning or prompt engineering, it’s genuine model pre-training, post-training, and large-scale systems research. The company secured substantial funding in early 2025 from top-tier institutional VCs and prominent angel investors, giving them a solid two-year runway and plans for another upcoming raise. The founding team includes senior engineers and researchers from Meta, Google, and AWS, each with deep experience training and deploying large-scale models. Their mission is simple but ambitious: to develop the next generation of language and multimodal foundation models and systems capable of general reasoning, memory, and efficient adaptation across domains. You’ll be joining a small, world-class team of researchers and engineers working collaboratively in a 5-day on-site environment at their central London offices. It’s a rare opportunity to get in early at a company that will define its own category once it comes out of stealth next year. The Role As a Research Engineer, you’ll play a key role in designing, training, and evaluating large-scale models and distributed systems. You’ll work closely with experienced colleagues who have scaled some of the world’s largest AI systems, helping to drive both the research and infrastructure that underpin next-generation foundation models. You’ll have a high degree of autonomy and ownership, contributing directly to experiments, training pipelines, and evaluation frameworks that will inform how the models evolve over time. What You’ll Work On • Designing and developing large-scale training pipelines for foundation models (language and multimodal)., • Building and scaling distributed training infrastructure across multi-GPU and multi-node environments., • Exploring pre-training and post-training techniques, including SFT, RLHF, RLAIF, and DPO., • Designing and automating evaluation systems to measure reasoning, alignment, and robustness., • Working closely with core research and systems teams to optimise training efficiency, data throughput, and inference performance., • Contributing to internal tooling for data curation, versioning, and reproducible experimentation. You’ll Bring • Strong experience with LLMs, generative, or multimodal models, ideally involving large-scale training or evaluation., • Hands-on fluency in PyTorch, JAX, or TensorFlow, with experience in DeepSpeed, Megatron, FSDP, or similar., • Understanding of distributed training, parallelism strategies, and scaling laws., • Familiarity with post-training methods (RLHF, RLAIF, DPO, or alignment optimisation)., • Solid software engineering fundamentals - version control, CI/CD, and production-grade model deployment., • Experience working in cloud-native or high-performance compute environments (AWS, GCP, Docker, Kubernetes)., • A curiosity for foundational research and a desire to work on systems that push beyond the current limits of LLMs. Why Join • Be part of a core founding technical team shaping a next-generation foundation model company., • Collaborate with world-class engineers and researchers from top AI labs., • Work on true pre-training and large-scale systems, not just fine-tuning., • Competitive salary up to £160k, plus meaningful equity., • Operate in a fast-paced, research-driven environment where ideas move quickly from prototype to production. 📍 Location: Central London 💰 Compensation: Up to £160k + equity 🏢 Work setup: 5 days per week on-site If you’re motivated by deep technical challenges and want to work on real model training and architecture design, not just incremental fine-tuning, this is one worth a conversation. To find out more, send a quick message to Jamie Forgan or apply directly with your CV, and we’ll arrange a confidential chat to run you through the details.
Page 11