Principal Machine Learning Engineer
Department:
Job Summary
About the role
Candidates must be eligible for Developed Vetting (DV) clearance.
More information about security clearance is available here: senior roles compensation is tailored to the scope of the role and the specific experience a candidate brings. Because that can vary significantly we do not always publish a fixed salary band. We will however discuss compensation openly at the first stage of the process and can share an indicative range before either side invests significant time.
- Technical authority: take technical ownership of our most demanding ML projects set the approach and make the decisions that matter.
- Stay hands-on at depth: design build train and debug yourself. This is a senior engineering role not an advisory one.
- Train and adapt large models: lead our large model work including pre-training fine-tuning distributed training and evaluation at scale.
- LLMs and agentic systems: bring practical expertise to projects across the team that build on LLMs and agents with an eye on latency cost and reliability in constrained settings.
- From data to the edge: make sure our solutions work end-to-end from data pipelines and training infrastructure through to deployment on edge hardware with evaluation that tells us what a model actually does.
- Represent Rowden externally: lead our technical presence at exercises trials and customer engagements and contribute technical expertise to bid writing.
- Raise the bar: set what good looks like with our craft and technical leads and mentor engineers through reviews pairing and example.
- Builder mindset: bias to action ownership over outcomes and comfort working through ambiguity.
- Proven delivery: you have made decisions that mattered and lived with the outcomes. You have led technical work into production or matured it to a high technology readiness level (TRL) ready for deployment.
- Deep ML and DL expertise: mastery across modern machine learning and deep learning with strong grounding in the maths and the ability to reason about trade-offs in production.
- Large model training: deep hands-on experience training and fine-tuning large models including distributed training across multi-GPU or multi-node environments and writing highly performant model training code.
- LLMs & agentic systems: a good level of knowledge and practical experience with LLM and agent-based systemsretrieval/RAG tool orchestration prompt optimisation and evaluation.
- Data and edge experience: practical experience with data engineering and deploying to edge or embedded hardware enough to design systems that work across the whole pipeline.
- Software development: excellent Python skills; experience with low-level languages such as Rust or C is desirable.
- Interpersonal skills: strong communicator who can influence without authority and bridge technical and non-technical audiences including senior customers.
- Education: MSc or equivalent experience required.
- Edge and embedded deployment: quantisation pruning distillation compilation and runtime optimisation for constrained hardware.
- Working with end users: experience at trials exercises or demonstrations or working directly with users in operational or other high-stakes settings.
- MLOps: reproducible pipelines model versioning CI/CD observability and automated evaluation.
- Education: PhD in AI/ML/CS or a related field.
- Distributed model training & infra: PyTorch DDP and FSDP Kubeflow Slurm Kubernetes and a large-scale training framework such as DeepSpeed Megatron or TorchTitan.
- LLM/Agent tooling: evaluation harnesses prompt optimisation agent frameworks.
- Inference and edge: TensorRT ONNX Runtime vLLM Nvidia Triton Inference Server KServe Nvidia Jetson (e.g. Jetson Thor).
- Data and platforms: Databricks Apache Spark MLflow SQL AWS GCP Docker.
Required Experience:
Staff IC
About Company
Rowden designs and builds systems, infrastructure, and applications to deliver mission advantage to those working to protect the security of the UK and its allies. We are setting a new standard in government technology provision: hyper-efficient engineering, better customer relationsh ... View more