Quick overview
Job description
About the Role
Join an early-stage industrial AI team building the multimodal AI core for wearable assistants used in field operations. You will own production vision-language systems end to end, from model orchestration and evaluation to deployment on real devices.
What You'll Do
Build and ship agentic vision-language pipelines that support visual reasoning and multi-step workflows.
Develop evaluation tools, capture failure modes, and create data and fine-tuning loops that improve model quality.
Adapt model serving to field connectivity, latency constraints, and edge hardware, including graceful fallbacks.
Create real-time voice and video interfaces, including conversational and proactive assistance modes.
Fine-tune open-source multimodal models for enterprise deployment, using approaches such as SFT, RLHF, and quantization.
What We're Looking For
Typically 2 to 6 years of experience building and shipping production VLM or multimodal AI systems, using Python and PyTorch.
Evidence of owning AI product outcomes, including agent loops with real tool use, model orchestration, evaluation, and data improvement workflows.
Applied experience fine-tuning and deploying models, rather than a focus solely on pretraining from scratch.
Experience with production vision-language systems, computer vision, and image or video reasoning; wearable, AR, industrial AI, or autonomous systems experience is valuable.
Familiarity with serving and optimizing models using tools such as vLLM, Triton, Ray Serve, ONNX, TensorRT, Hugging Face Transformers, and quantization methods.
A CS, machine learning, or engineering degree and a record of independently driving product work are preferred.
Compensation & Benefits
Base salary: $180,000 to $230,000 annually. Visa sponsorship is available.
Location
On-site in San Francisco, United States.