Integration Engineer, Clinical Data Path

a2z Radiology AI

2d•$140,000 - $180,000•Hybrid

About The Position

This role owns the clinical data path, ensuring that studies move from hospital systems into a2z Radiology AI, that results get back to the right systems, and that the entire process is proven to be working correctly. The company is building clinical AI that works alongside radiologists, with an FDA-cleared abdomen-pelvis CT triage device. This position is crucial because a cleared device only provides value when it functions seamlessly within existing radiologist workflows. The clinical data path involves software within hospital networks, cloud DICOMweb ingest, de-identification, tenancy, result delivery, and meeting the evidence requirements of hospital IT reviewers. Two gateway repositories require continuous integration, and this role will be the sole owner responsible for the end-to-end integration surface.

Requirements

Implemented a protocol from its specification, not just used a library that wrapped one.
Experience running software inside infrastructure where direct SSH access is not possible.
Instinct to widen the system boundary for debugging when encountering issues like a bare 502 with no application logs.
A fail-closed approach is a primary instinct when dealing with patient data.
Ability to write precise, traceable, and honest documentation suitable for hospital IT reviewers.
Protocol literacy, outside-in debugging skills, and fail-closed instincts are essential.
Python, FastAPI, pytest experience.
Experience with pynetdicom, pydicom.
Familiarity with DICOM PS3.15 and PS3.18 standards.
Experience with DICOMweb STOW-RS.
Experience with HL7 v2 over MLLP.
Experience with AWS, Terraform, Docker.
Experience with shell scripting and installers that function on locked-down hospital systems.

Nice To Haves

DICOM, PACS, VNA, or teleradiology systems experience.
HL7 interface engines experience.
Experience with pynetdicom, pydicom, Orthanc, dcm4che, or DCMTK.
Networking, embedded, or payments systems experience.
De-identification or privacy engineering experience.
Confidential computing or hardware attestation experience.

Responsibilities

Own on-premise DICOM ingest, including a pynetdicom C-STORE service that assembles studies, handles incomplete sender signals, applies backpressure, passes compressed transfer syntaxes, and spools safely during link drops.
Manage cloud ingest, specifically a DICOMweb STOW-RS receiver with bounded-memory multipart parsing, study-UID cross-checking, and exactly-once downstream firing per request.
Oversee tenancy and privacy at the network boundary, including customer resolution without silent defaults, adherence to DICOM PS3.15 confidentiality rules, fail-closed checks before data leaves the customer network, and managing a re-identification store under customer-held keys.
Manage outbound data delivery, including JSON callbacks, dictation-system autotext, and HL7 v2 ORU^R01 over MLLP, tailored to the customer's interface engine.
Ensure visibility and correctness of the data path, including per-study tracing, three-component integration health monitoring, configuration-drift checks, and CI for both gateway repositories.
Provide precise technical information for partner reviews regarding transport, retention, incident handling, and system capabilities.
Debug complex integration issues, including those that manifest differently than error messages suggest, such as large request bodies causing OOM errors before logging, or incorrect allowlist configurations leading to transfer failures.