AI Faithfulness Engineer

About the role

Help companies ship AI answers that stay true to the source material they were given. Through Nebulai's marketplace, you join teams running AI assistants in production: support bots, document summarizers, research tools, and internal knowledge search. You find the spots where the model adds facts, drops caveats, or changes numbers, build checks that catch unsupported statements before users see them, and tune retrieval, prompts, and output rules so answers say only what the documents support.

Requirements

• Have shipped an LLM feature where wrong or invented answers had a real cost
• Can trace a hallucinated claim back to its cause in retrieval, prompting, or the model
• Have measured faithfulness or answer support with real eval sets
• Work well with product, data, and engineering teams

Preferred qualifications

• Experience in healthcare, finance, legal, or other regulated industries
• Experience across on-prem and multi-cloud stacks without favoring a vendor

Engagement

Contract AI faithfulness and evaluation work via Nebulai. Apply at https://nebulai.app