LLM Data Engineer- Citizen Only

Lightning Minds Inc.

United States

about 1 month ago

Apply on www.dice.com

Prepare application

Role details

Contract type

Permanent contract

Employment type

Full-time (> 32 hours)

Working hours

Regular working hours

Job source

www.dice.com

Tech stack

Application Programming Interfaces (APIs) Artificial Intelligence Amazon Web Services Amazon S3 Audit Trail Clinical Data Repository Continuous Integration Data as a Services Information Engineering Data Files Dicom Identity and Access Management

+11 more

SQL Databases Fast Healthcare Interoperability Resources Large Language Models Apache Spark Amazon Virtual Private Cloud (VPC) Containerization AWS Glue Health Level Seven International AWS Data Analytics Terraform Docker

Job description

Build ingestion and normalization pipelines for three distinct modality families: DICOM radiology studies, whole-slide pathology images, and tabular EHR extracts (labs, vitals, encounters, medication administration records).
Design the canonical benchmark record format, which is the intermediate representation that every task configuration and every model adapter reads from, so that full EHR record and image only variants of the same task are provably drawing from the same underlying case.
Solve the modality packaging problem: gigapixel pathology slides and multiseries radiology studies must be reduced to payloads that fit inside third-party API limits (48 images, 20 MB) without silently destroying diagnostic signal. You will build the tiling, region selection, downsampling, and compression strategies, and the provenance metadata that records exactly what was sent so results remain reproducible and defensible.
Construct the cohort and label pipelines behind the prediction tasks: timewindowed feature assembly for sepsis onset, survival horizon calculation for days-to-death, and longitudinal series construction for lab value trends.
Build the results store and analysis layer per-run, per-model, per-task scored outputs.
Enforce PHI handling discipline end to end: encryption at rest and in transit, least-privilege access, audit logging, deidentification where required, and clear boundaries around what leaves the VPC when a payload goes to a third-party API.

Requirements

AWS, Data Sets, data sources, data services with in AWS
Strong AI & LLM
Recent healthcare industry exp (HIPAA, hl7, etc)
LinkedIn Page
Strong communication
US Natural Citizen considered first and among all candidates., * Data engineering: Production-grade pipeline code, not notebooks. Strong testing discipline with developing deterministic, idempotent, re-runnable jobs.
AWS data stack: Deep hands-on experience with S3 (layout design, lifecycle policies, storage class economics at image scale), AWS Glue and/or Spark on EMR, Athena, Step Functions, Lambda, and AWS Batch.
SQL and data modeling: Complex temporal joins, point-in-time correctness, and the discipline to avoid label leakage in time-series clinical data.
Data quality and lineage: Validation frameworks, schema enforcement, and versioned datasets.
AWS security and governance: IAM policy design, KMS, VPC endpoints and PrivateLink, and working inside a HIPAA-eligible architecture with a BAA in place. Desirable Skills
Healthcare data formats and standards: DICOM, FHIR, HL7v2, OMOP CDM, and the practical realities of clinical coding (ICD, LOINC, RxNorm, SNOMED).
Medical imaging handling: pydicom, OpenSlide, and the basics of WSI pyramid structure and tiling.
Infrastructure as code (Terraform) development.
Containerization and CI/CD: Docker, ECR, and a mainstream CI system.
Enough familiarity with LLM APIs and multimodal payload construction. Nice to Have
Prior work on clinical prediction models or healthcare ML datasets (MIMIC, eICU, or equivalent institutional data).
Experience with AWS HealthImaging or HealthLake.
Familiarity with de-identification tooling and re-identification risk assessment.