Senior Edge Dialog AI Engineer - On-Device Safety & NLU

BrightAI; Palo Alto (CA); Hybrid; USD 170,000 - 210,000; Full time

Job summary

BrightAI in Palo Alto, CA is hiring a Sr. AI Engineer – Edge Dialog Systems to own the on-device conversational AI powering our industrial safety wearable.

You will design end-to-end dialog pipelines, optimize memory and latency, and enforce deterministic safety guards while keeping computation on-device. You’ll work at the intersection of NLP, small language models, and embedded software, building multi-turn dialog/state management with local-first inference and data-driven configuration to

Qualifications

• 5+ years of ML/AI experience with NLP/LLMs or conversational AI.
• Experience with embeddings and semantic similarity, RAG, and evaluation.
• Strong Python, pytest, Git, and CI discipline.

Responsibilities

• Own end-to-end on-device dialog pipeline and safety gating.
• Maintain deterministic safety layer around the language model.
• Run and optimize SLM inference on-device within latency and memory budgets.
• Extend zero-shot configuration to support new device commands and procedures.
• Collaborate with product, firmware, and cloud teams.

Skills

ML/AI experience

NLP / LLMs

Prompting

Structured output

Tool calling

Evaluation

RAG

Embeddings

Python

pytest

Git

CI

Edge NLP

Dialog state mgmt

Local-first inference

On-device ML

Linux

Docker

adb

systemd

ONNX

quantization

embedded targets

automation

problem-solving

Education

AI/ML/NLP background

Embedded software experience

Tools

ONNX

onnxruntime

Job description

BrightAI in Palo Alto, CA is hiring a Sr. AI Engineer – Edge Dialog Systems to own the on-device conversational AI powering our industrial safety wearable.

You will design end-to-end dialog pipelines, optimize memory and latency, and enforce deterministic safety guards while keeping computation on-device. You’ll work at the intersection of NLP, small language models, and embedded software, building multi-turn dialog/state management with local-first inference and data-driven configuration to