Senior Machine Learning Engineer (ML Platform)
Posted Updated
Summary
A hands-on platform role owning the ML monorepo for DeepHealth's X-ray/CT/MRI team: dataset definitions, training/evaluation loops, CI tooling, experiment tracking, and production inference servers on GCP and HPC. Also leads the team's agentic coding strategy. Core stack: Python, PyTorch, GCP, containers.
- The Senior Machine Learning Engineer, ML Platform is responsible for building, running, and evolving the ML monorepo used by the X-ray, CT, and MRI machine learning group at DeepHealth. The monorepo covers the full path from experimentation to product: dataset and sample definitions, training and evaluation code, experiment tracking, model export, and the inference servers that run in production
- Everyone on the team contributes to the monorepo; this role is the single point of accountability for it. This hands-on role owns the architecture, tooling, and engineering standards of that codebase for a distributed team of about twenty machine learning engineers and scientists, working with model developers and software engineers to make the team’s work fast, better, and easier to audit
- It also leads the team’s agentic coding strategy, shaping how AI coding agents are used across the codebase. The role reports to the Leader of the ML for X-ray/CT/MRI group
- Own the architecture and roadmap of the ML Monorepo:
- Module boundaries, shared abstractions for datasets and samples, training and evaluation loops, configuration, and experiment tracking
- Build and run the engineering tooling around the codebase:
- Continuous integration, testing, dependency management, containers, artifact and model registries, and GPU job scheduling on cloud (GCP) and HPC infrastructure
- Own the path from trained checkpoint to deployable model:
- Model packaging, inference servers with enforced API contracts, reproducibility, and performance and memory budgets on target hardware
- Define and enforce engineering standards (code review, testing, documentation) in a way that accelerates research rather than slowing it down
- Ensure traceability of data, code, and model versions across experiments and releases
- Investigate and resolve performance, reliability, and usability issues in the stack raised by the team and reduce their recurrence through shared tooling
- Drive the team’s agentic coding strategy: define how AI coding agents are used in the monorepo, set up the tooling, guardrails, and conventions that make them effective, and help colleagues adopt them
- Review code, mentor colleagues on engineering practices, and communicate technical decisions and trade-offs clearly
- Experience taking models to production: model packaging, inference serving, and performance and reliability constraints (required)
- Ability to work autonomously in a distributed, multi-site team, with clear verbal and written communication in English (required)
- Solid PyTorch knowledge and practical experience with model training and profiling (required)
- Contributions to open-source ML tooling (preferred)
- Experience running ML workloads on a cloud platform, ideally GCP (Vertex AI, Artifact Registry, IAM, storage), and with containerization (required)
- HPC and Slurm experience (preferred)
- 5+ years of software engineering experience, including a significant portion building infrastructure or tooling used by machine learning teams, such as an ML platform or a training framework (required)
- Strong Python skills and software engineering fundamentals: API design, testing, refactoring, and attention to developer experience (required)
- Master’s degree in computer science, engineering, mathematics, or a related field, or equivalent practical experience (required)
- Medical imaging experience (DICOM, 3D volumes, large images) (preferred)
- Ability to work closely with researchers, translate recurring pain points into shared tooling, and push back constructively when needed (required)
- Follows all local, regional and country laws concerning employment
- Follows data privacy, compliance, safety and confidentiality standards at all times
- Adapts and is willing to learn new tasks, methods, and systems
- Follows all DeepHealth policies and procedures
- Reports to work regularly as scheduled; consistently punctual with respect to working hours, meal and rest breaks, and maintains satisfactory personal attendance in accordance with DeepHealth guidelines
- Demonstrates initiative, personal awareness, professionalism and integrity, and exercises confidentiality in all areas of performance
- Completes job responsibilities in a quality and timely manner
- Practices universal safety precautions
- Communicates, cooperates, and consistently functions professionally and harmoniously with all levels of supervision, co-workers, visitors, and vendors
- Promotes good public relations on the phone and in person