About the Opportunity
This role is the single point of accountability for a medical imaging backend: its correctness, its behaviour in production, and the written record behind each technical decision. You will own the clinical data, the workflow around it, and the accuracy of the numbers that reach a report. The work includes the unglamorous parts, such as chasing failures that only appear under load and keeping a retiring service running until it is fully replaced. Machine learning models are built by a separate data science team, so your focus is the backend that serves them.
What You'll Do
- Own the Python/FastAPI backend end to end, including HTTP APIs, background workers, and production behaviour.
- Own the PostgreSQL data model and its evolution, covering schema design, migrations, concurrency control, and query performance.
- Build asynchronous, event-driven pipelines for medical image processing, from ingestion and de-identification to inference submission and result handling, with reliable retry and timeout behaviour.
- Integrate with an internal platform for authentication, GPU inference, and usage metering, keep the product resilient when it is unavailable, and enforce multi-tenant data isolation and patient data privacy.
- Agree the product-to-model contract with the data science team and define API contracts with the frontend engineers who consume them.
- Maintain the architecture decision record, lead design reviews, and support the legacy TypeScript service until it is retired.
Core Essentials
- 5+ years of professional backend engineering, including production ownership of a system you designed.
- Deep backend fundamentals in any language (Python, Go, Java, Kotlin, C#, TypeScript or Rust), and willingness to become effective in Python, FastAPI, and an asynchronous codebase.
- A strong instinct for concurrency hazards such as race conditions, lost updates, and messages processed twice.
- Strong PostgreSQL and SQL: data modelling, transactions and concurrency control, indexing, and migration practice.
- Experience with message queues and event-driven architecture, including idempotent retries, dead-lettering, and timeout handling, and with integrating external services designed for failure.
- Disciplined engineering practice (static typing, automated testing, code review, CI/CD, Docker) and productive use of AI coding assistants.
- Ability to work independently, set technical direction, and raise concerns about a requirement before implementing it; comfortable working in both Chinese and English.
Bonus Skills
- Hands-on Google Cloud Platform experience.
- Medical imaging experience, particularly DICOM and libraries such as pydicom, or a background in regulated healthcare software handling protected health information.
- Multi-tenant SaaS experience, especially database-enforced isolation such as row-level security, or products that also ship on-premise.
- Numerical or geometric computing (NumPy, computational geometry) and working familiarity with React and TypeScript.
