About this role
Principal AI Engineer: Voice, Memory & Clinical Safety (Fractional)
- $70k – $140k • 0.05% – 0.25%
- Remote (Canada • United States +2)
- 5 years of exp
- Contract
Posted: today • Recruiter recently active
Hires remotely in
Canada -
United States -
United Kingdom -
Western Europe
Remote Work Policy Remote only
Company Location
San Diego • San Diego County
Visa Sponsorship Not Available
Preferred Timezones Pacific Time
Collaboration Hours 8:00 AM - 12:00 PM Pacific Time
RelocationNot Allowed
Skills
-
Python
-
Machine Learning
-
PostgreSQL
-
WebRTC
-
TypeScript
-
Docker
-
AWS
-
NLP
-
Generative AI
-
Large Language Models (LLMs)
-
Retrieval-Augmented Generation (RAG)
-
Amazon Bedrock
-
About NuVee and Koa
-
NuVee is building Koa, an AI care system designed to support people using GLP-1 medications between clinical visits. Koa is live on iOS and preparing for her first real-world patient pilot.
-
What You Will Own
-
You will work directly with our CTO, physician founder, and engineering team to strengthen Koa’s intelligence, reliability, and safety.
Your initial responsibilities will include:
- Audit the current voice, memory, context, reasoning, and safety architecture.
- Improve longitudinal memory, temporal pattern detection, and user-state management.
- Strengthen real-time voice orchestration, interruptions, latency, and context handoffs.
- Build an evaluation system that tests each safety function separately:
- Detect the relevant signal.
- Ask the appropriate confirmation question.
- Select the correct escalation.
- Complete the handoff.
- Monitor the later outcome.
- Create regression tests for memory retrieval, pattern recognition, intervention selection, escalation, and follow-up.
- Instrument model performance, latency, retrieval quality, failure rates, token and voice costs, and safety behavior.
- Identify when a failure is caused by prompting, architecture, data, workflow, or product design.
- Improve the existing system without proposing an unnecessary ground-up rewrite.