AI Engineer, Document and Decision Steps

At worklist, AI is a step, not the architecture. Each AI step has typed inputs and outputs, shows its sources, and reports confidence as Confident or Check this. You'll build the extraction, classification and drafting steps, and the evaluation that proves they're good enough to run on their own.

What you'll do

• Build extraction for faxes, referrals, EOBs and medical records
• Build classification of denials and drafting of appeal letters from evidence and payer policy
• Design evaluation sets from real (de-identified) cases and track quality per step
• Make every output explainable: sources, rationale and the mode it ran in
• Keep PHI out of logs, prompts' telemetry and test fixtures

What you'll bring

• 3+ years shipping machine learning or LLM features to production
• Experience with document understanding and structured extraction
• A rigorous approach to evaluation: test sets, regressions, confidence calibration
• Python or TypeScript, and comfort working alongside a PHP backend

Nice to have

• Healthcare documents: CMS-1500, UB-04, EOBs, prior authorization forms
• Retrieval over policy documents (LCDs, payer medical policies)

About worklist

worklist runs healthcare operations as workflows, from intake to payment, and turns everything automation can't finish into structured work for a person. We're a small team building it carefully, starting with denial recovery for medical equipment suppliers.