Senior Data Pipeline Engineer (DPE)

1. Role Summary

The Senior Data Pipeline Engineer (DPE ) owns the data pipelines that run inside our hospital partners. Oncoshot’s architecture keeps patient-level data within the institution: at each hospital, a pipeline draws records from clinical systems, converts clinical documents into machine-readable text, de-identifies them, and passes only de-identified content to the Oncoshot platform for structured extraction. The DPE designs, builds, deploys and maintains those pipelines end to end, and is accountable for their reliability, throughput and measured quality at every site.

This is a senior, hands-on delivery role. The DPE works close to the data and close to the hospital environment, engineering to the constraints that come with clinical data — restricted networks, infrastructure we do not own, and the absolute requirement that no identifiable data leaves the institution. Success in this role is measured by pipelines that run unattended, quality that is proven with numbers rather than asserted, and new institutions that come online faster and more repeatably each time.

2. Key Responsibilities

• Own the pipelines end to end: Design, build and maintain the full path from hospital clinical systems through document conversion and de-identification to loading into the Oncoshot platform — including resumable and idempotent processing, backfills, per-document lineage and reconciliation against the source system.

• Deploy into environments we do not control: Own how the pipelines are packaged, distributed, installed, upgraded and rolled back — reproducibly, across on-premises hospital infrastructure and cloud environments, including network-restricted and offline sites.

• Quantify the quality of every stage: Establish and publish the right measure for each step — coverage and reconciliation for data extraction and ETL, error rates and field-level accuracy for document conversion, and recall-weighted measures (F2) with an explicit precision floor for de-identification, where a missed identifier and an over-redacted clinical detail are not equivalent errors.

• Own the evaluation harness and release gates: Maintain gold-standard datasets, per-site quality baselines and the automated regression gates that prevent a change from shipping when quality moves in the wrong direction.

• Run and support production: Monitor pipelines across sites and diagnose issues on data that cannot be inspected — using privacy-preserving telemetry, reproduction on synthetic data, remote support alongside hospital operators, and on-site troubleshooting where a problem requires presence.

• Commission new institutions: Bring new hospital sites live: assess their systems and constraints, adapt the pipeline, validate quality against agreed thresholds, and hand over runbooks and operating procedures that hospital staff can follow without engineering support.

• Engineer within security and compliance constraints: Handle credentials and secrets safely on third-party infrastructure, keep identifiers out of logs and support tickets, produce the evidence expected by an ISO 27001 programme, and respect data-residency and cross-border transfer obligations.

• Partner across the organisation: Work with hospital IT, clinical and operations stakeholders, and with Oncoshot’s platform, machine-learning and commercial teams — translating pipeline performance into terms each audience can act on.

3. Preferred Qualifications

• Bachelor’s or Master’s degree in Computer Science, Software/Data Engineering, Information Systems, Biomedical Informatics, or a related field.

• 5+ years of professional software or data engineering experience, including at least 3 years building and operating production data pipelines at senior or lead level, with deep Python.

• Strong Linux engineering: containers and orchestration, packaging and distribution, and operating in restricted, offline or on-premises environments.

• Cloud infrastructure experience on at least one of AWS, Azure or GCP, managed as code, with the judgement to recognise that on-premises deployment is a different discipline.

• Experience running machine-learning models in production — for example OCR, named-entity recognition or document AI — covering model versioning, CPU/GPU trade-offs and performance within fixed compute and memory budgets.

• Demonstrated evaluation literacy: precision, recall, F1 and F2, error analysis, gold-standard datasets, and regression testing of models and rule sets.

• Experience in healthcare, clinical research, life sciences or health-data domains is strongly preferred (e.g., EMR/EHR, HL7/FHIR, DICOM, ICD/SNOMED coding, clinical documents, de-identification).

• Familiarity with privacy and information-security frameworks relevant to health data — PDPA, GDPR, HIPAA — and with ISO 27001 in practice.

• Comfortable working within an Agile/Scrum delivery model.

4. Core Competencies

• Delivery ownership: Takes a capability from design through to running reliably in production, without handing off the difficult parts.

• Measurement rigour: Chooses the right metric for each stage, understands when a symmetric measure such as F1 is the wrong choice, and lets the numbers drive decisions.

• Debugging under constraint: Diagnoses failures in systems that cannot be observed directly, and designs for supportability from the outset.

• Designing for other people’s environments: Builds for reproducibility, restricted infrastructure and non-engineer operators rather than for a developer workstation.

• Security and privacy instinct: Treats the boundary between identifiable and de-identified data as absolute, and reasons about risk before it becomes an incident.

• Communication across audiences: Explains technical quality and trade-offs credibly to clinical, commercial and engineering stakeholders.

5. Success Measures (First 3 Months during Probation Period)

• Assumed full ownership of the existing pipelines and demonstrated the ability to operate, support and troubleshoot them independently.

• Published a measured quality baseline for each pipeline stage, with the metrics and thresholds agreed with the COO and CTO.

• Delivered at least one substantive improvement or new site onboarding, released through a repeatable process with quality gates in place.

6. Working Model

• Probation period: First 3 month probation period and conversion to permanent staff after passing the probation.

• Reporting line: Reports to the Chief Operating Officer, with the Chief Technology Officer consulting on and reviewing technical design and delivery.

• Location & arrangement: Hybrid working, based in Singapore, with a mix of on-site collaboration and remote work.

• Hospital engagement: Periodic on-site presence at hospitals in Singapore and occasional travel to sites in the region, to commission new institutions and support existing ones. Hospital maintenance windows occasionally require out-of-hours work.

• Eligibility: Open to Singapore Citizens and Permanent Residents only.

• Team structure: Works cross-functionally with hospital partners and with Oncoshot’s platform, machine-learning and operations teams.

• Cadence: Operates within an Agile/Scrum delivery model, participating in sprint planning, refinement and reviews.

7. Salary Range

Indicative monthly gross salary: SGD6,000 – SGD8,000 (exclude CPF for PR/Singaporean and performance bonus/benefits/Employee Equity Share), commensurate with experience and seniority.