AI/LLM Engineer - Production Claims AI

Codoxo; Duluth (GA); Hybrid; USD 130,000 - 190,000; Full time

Benefits offered by this job

Health, dental, and vision insurance

Unlimited PTO

Annual home office stipend

401K match

Job summary

Codoxo in the United States is seeking an AI/LLM Engineer to design, ship, and operate LLM-powered features that accelerate claim audits and SIU investigations—retrieval and QA over case documents, structured extraction from claims and medical records, and investigator-facing summaries.

You will work with Python and Django on AWS, using Bedrock for inference, OpenSearch for vector search, and Celery for asynchronous processing, delivering end-to-end features to fraud investigators at national

Qualifications

• 3+ years of professional Python experience.
• 2+ years building with LLMs in production.
• Hands-on RAG: embeddings, vector search, chunking, diagnostics.
• Experience with evaluation: metrics, eval sets, true/false results.
• Working knowledge of AWS serverless and PHI handling.

Responsibilities

• Build and ship LLM features for document QA, extraction, and summarization.
• Own retrieval pipeline end to end: ingestion, filtering, search, and citations.
• Develop evaluation systems: golden sets, regression suites, rubric grading.
• Treat prompts as engineering artifacts with versioning and reviews.
• Design structured output schemas and handle refusals gracefully.
• Work on serverless AWS: Bedrock, Lambda, S3, API Gateway, Step Functions.
• Monitor cost and latency; implement token caching and routing.
• Collaborate with investigators and product to define evaluation criteria.

Skills

Python

LLMs in production

RAG experience

Evaluation practice

PHI data handling

Tools

OpenSearch

Bedrock

Lambda

S3

API Gateway

Step Functions

Celery

Job description

Codoxo in the United States is seeking an AI/LLM Engineer to design, ship, and operate LLM-powered features that accelerate claim audits and SIU investigations—retrieval and QA over case documents, structured extraction from claims and medical records, and investigator-facing summaries.

You will work with Python and Django on AWS, using Bedrock for inference, OpenSearch for vector search, and Celery for asynchronous processing, delivering end-to-end features to fraud investigators at national