Eventual Computing
Research Engineer, Multimodal Data
$150K–$300K
San Francisco · Full-time · Posted Apr 29, 2026
About the job
Own and ship visual-understanding models and pipelines that make large video and sensor-data collections searchable and produce curated training datasets for Physical AI customers.
What you'll do
- Set the visual-understanding roadmap, select models, and deploy corpus-scale inference.
- Train, fine-tune, and evaluate multimodal and vision models; reduce annotation costs through model and pipeline optimization.
- Build queryable, quality-instrumented datasets and work with partner researchers and internal data teams.
What they ask for
- Strong familiarity with modern vision and multimodal models, including convolution nets, VLMs, VQA, and embeddings.
- Experience running models at scale on real video and sensor data, ideally for perception tasks.
- Perception-team background in self-driving, robotics, or visual-data companies, or equivalent depth from a research lab.
- Comfort with cloud infrastructure and large-scale data processing; shipped jobs involving thousands of GPU-hours of video.
Nice to have
- Experience training and evaluating vision or multimodal models.
- Research background, experience with embeddings or retrieval at scale, or experience designing labeling taxonomies or annotation programs.
Pay
$150K – $300K • Offers Equity
Tools
VLMs, VQA models, embedding models, CNNs, cloud infrastructure
AI focus
Applied vision and multimodal ML research: train, fine-tune, evaluate, and deploy models for corpus-scale video and sensor-data annotation, search, and training-data curation.
Model-written summary · the original posting is authoritative