The role
- COMP
- $180K - $230K
- EQUITY
- 0.25-0.75%
- LOCATION
- San Francisco
- WORKPLACE
- On-site
- EXPERIENCE
- 0 - 3 years
- VISA
- None, Visa transfers
- STACK
- Python, PyTorch, Langchain, Docker
- INDUSTRY
- AI, Hardware, Robotics, Enterprise, Defense, B2B
/roles — ROLE_380
Founding ML Researcher
Two-founder startup running multimodal AI on wearable glasses to guide technicians in heavy industry
The company
Wearable AI company putting real-time audio and visual assistance on smart eyewear to guide workers through complex industrial assembly and inspection work.
- STAGE
- Series A-stage
- FUNDING
- $5M raised
- TEAM
- founding team of <10
- FOUNDED
- 2026
JD — the work
About the role
You would join a two-person founding team as one of its first engineers and build the core AI behind smart glasses used by field workers at data centers, in aerospace, and on power grids. The scope is the whole multimodal stack from the start: production pipelines for vision-language models, evaluation frameworks, on-device inference, and real-time voice and video assistance running on real hardware. It suits a scrappy builder who has put agent-based or multimodal AI in front of real users, beyond demos, and cares more about computer vision in legacy industries than chasing LLM trends.
What you'll do
- Deploy vision-language agents on the glasses that reason step by step and call tools during real inspections and procedures
- Run the evaluation and data loop, capturing failure modes and using customer data to fine-tune better models each release
- Handle edge inference and changing connectivity with orchestration, runtime tuning, and graceful fallbacks
- Create real-time voice and video assistants for the glasses, from spoken conversation to proactive alerts, tuned to each user type
- Lead multimodal training, adapting open-source models with SFT, RLHF, and quantization for customers who deploy on-premise
What they're looking for
- 0 to 3 years of experience building and shipping AI systems
- Proof you have delivered agent-based or multimodal AI into actual users' hands, beyond prototypes
- Strong Python and PyTorch skills for model work
- Genuine enthusiasm for wearables, computer vision, and applied AI in traditional industries
- A scrappy, high-ownership approach and readiness to work in person in San Francisco