AI Ticket Deflection Automation Engineer

Tuvex Solutions

Remote · $130k – $160k / year · Full-time · Posted 4 months ago

This role is no longer accepting applications

Don't miss the next role like this

Create a free profile with your skills. We match you to new AI roles as they open, and companies hiring can contact you directly.

Create free profile, 2 minutes

Similar open roles you can apply to now:

See all open AI jobs

This role is no longer accepting applications

Don't miss the next role like this

Create a free profile with your skills. We match you to new AI roles as they open, and companies hiring can contact you directly.

Create free profile, 2 minutes

Similar open roles you can apply to now:

See all open AI jobs

About the role

Our clients' support teams are buried. The median company on our platform receives 3,400 support tickets per month, of which 38% are questions already answered in their documentation. Every one of those tickets costs $8-15 in agent time before it's resolved. Deflection — answering the question automatically before it becomes a ticket — is the highest-leverage intervention we can ship. You'll build the deflection engine that sits in front of the ticketing system: a RAG-based system that ingests customer documentation, matches incoming questions to existing answers, generates accurate responses with source citations, and routes to human agents when confidence is below threshold. The hard part isn't the happy path — it's the fallback logic: knowing when to escalate, how to handle questions that span multiple documents, and how to prevent confident wrong answers that would be worse than no answer. The platform team is eight engineers. You'll own the deflection system end-to-end, from document ingestion to answer quality monitoring. You'll set the quality standards and have veto power over any change that risks increasing the confident-wrong-answer rate.

Fully remote.

Responsibilities

Build a RAG-based ticket deflection system for customer support automation
Design document ingestion, chunking, and embedding pipelines for client knowledge bases
Implement confidence scoring and escalation logic for uncertain answers
Create answer quality monitoring that catches confident wrong responses
Build feedback loops from agent corrections into training and retrieval improvement
Define quality SLAs and monitor deflection rate vs. false deflection rate trade-offs

Requirements

5+ years building production NLP or conversational AI systems
Deep experience with RAG architectures, retrieval systems, and LLM orchestration
Strong understanding of fallback logic and uncertainty quantification in LLM systems
Python and FastAPI proficiency for backend service development
Track record shipping customer-facing AI systems with measurable business impact