AI Engineer, Medical Vision Language Models

Lumexa Imaging • San Francisco, CA

2d • Remote

About The Position

Lumexa Imaging is hiring an early-career AI Engineer to build, adapt, and deploy vision language models (VLMs) and imaging foundation models for clinical radiology. This is a hands-on, build-focused role centered on model development and deployment. The engineer will use agentic coding tools, including Claude Code and OpenAI Codex, to turn radiologists’ ideas into validated, deployed models. Initial priorities include building clinical LLM pipelines for prior report summarization, scanned document summarization, and relevant prior selection, using infrastructure shared with VLM work.

The role provides access to a large and growing repository of clinical imaging and reporting data. The engineer will use real-world, domain-specific data to develop, adapt, and evaluate AI models in accordance with applicable privacy, security, and governance requirements. The position works closely with the Chief Clinical AI Officer, the Clinical Applications team, and Infrastructure and Information Security.

Requirements

• Hands-on experience fine-tuning transformer models in PyTorch (vision, multimodal, or language) using Hugging Face, MONAI, or similar, through internships, research, open-source work, or projects.
• Working knowledge of embeddings and vector search.
• Experience with open VLMs such as Qwen-VL or LLaVA.
• Daily use of agentic coding tools such as Claude Code and OpenAI Codex.
• Familiarity with Docker and GPU-based model serving.
• Comfort working with real-world data and a reproducible approach to experiments.
• Degree in Computer Science, Biomedical Engineering, or a related field, or equivalent practical experience.
• 0 to 3 years of experience; internships, research, or projects count.

Nice To Haves

• Hands-on experience with vision encoders and contrastive models such as CLIP, SigLIP, or JEPA.
• Exposure to medical VLMs, imaging foundation models, or self-supervised methods such as DINOv2 or MAE.
• Experience building MCP servers in Python or TypeScript.
• Familiarity with medical imaging formats such as DICOM, HL7, and FHIR, and HIPAA/PHI handling.
• Awareness of MLOps tooling such as MLflow, Weights & Biases, and model registries.
• Awareness of FDA considerations for clinical AI, including 510(k) and PCCP.

Responsibilities

• Fine-tune and adapt open vision language models and imaging foundation models for report drafting, impression generation, finding detection, and prior comparison, using parameter-efficient methods such as LoRA and QLoRA as well as full fine-tuning under senior guidance.
• Work with vision encoders, contrastive image-text models, and self-supervised approaches; build and maintain embedding pipelines for retrieval, cohort discovery, case similarity search, and drift detection.
• Curate and version datasets from internal PACS and reporting systems, including de-identification of DICOM headers, burned-in pixels, and paired reports.
• Use Claude Code and OpenAI Codex daily to build, test, and maintain code; maintain project context files such as CLAUDE.md and AGENTS.md; and help build and run MCP servers for internal system access.
• Deploy models using containerized inference and GPU serving, support CI/CD pipelines, and monitor production models for drift and performance.
• Build automated evaluation harnesses to assess clinical accuracy, negation, laterality, and hallucinations against radiologist ground truth.
• Build agentic and LLM-driven tools for operational workflows, including converting unstructured requests into RIS scheduling actions and supporting intelligent operations dashboards.
• Build clinical LLM pipelines for prior report summarization, scanned document summarization, and relevant prior selection.

What This Job Offers

• Job Type: Full-time
• Career Level: Entry Level