Hadrian
Machine Learning Engineer - Multimodal
Los Angeles, CAFrom $160kfulltimemidAdded today
About this role
Hadrian seeks a Senior Machine Learning Engineer to develop multimodal vision-language models that interpret manufacturing documentation and design intent, replacing traditional OCR pipelines with unified architectures for autonomous factory automation in aerospace and defense.
What you'll do
- Research and deploy multimodal models understanding manufacturing documentation across images and text
- Build annotation tooling, implement active learning loops, and engineer synthetic data augmentation strategies
- Develop evaluation frameworks that measure system behavior and real-world user impact beyond benchmark metrics
- Collaborate with ML team to shape technical and product roadmaps for the AI platform
- Optimize model accuracy to maximize time savings at manufacturing scale
What they're looking for
- Deep learning (5-8 years professional experience)
- Multimodal models (vision + language, 2+ years)
- Python and PyTorch (custom training loops, loss functions, data loaders)
- Production ML deployment and model monitoring
- CAD/CAM data processing
- Computational geometry
- Active learning and data augmentation
- Evaluation framework design
Benefits
- Medical, dental, vision, and life insurance
- Opportunity to impact aerospace and defense manufacturing at scale
- High-ownership startup environment
- Work with cutting-edge multimodal AI systems
- Competitive compensation and equity
Opens the official application on the employer’s site. No login required.
Hadrian
Hadrian builds aerospace and defense manufacturing systems, offering enterprise software platforms, advanced tooling design, and highly automated production capabilities for the sector. The company is hiring full stack engineers, manufacturing and tooling specialists, infrastructure and identity management experts, and workforce systems architects to support its rapid scaling.
View all jobs at HadrianLikely interview questions
- Describe a multimodal model you've built from scratch—what architectural choices did you make to fuse image and text, and how did you validate performance?
- Walk us through a time you owned a model in production. How did you monitor drift, and what were your strategies for maintaining accuracy?