Principal Machine Learning Engineer (Reconstruction / Quantitative Imaging)

Principal Machine Learning Engineer (Reconstruction / Quantitative Imaging) at Midjourney — San Francisco, CA, US

  • Company: Midjourney
  • Location: San Francisco, CA, US
  • Employment type: Full-time
  • Posted: 2026-06-30

About this role

WHAT YOU’LL DO

  • Partner with medical image reconstruction scientists / engineers to build ML components that improve reconstruction quality, speed, robustness, or quantitative accuracy.
  • Define training/evaluation pipelines, datasets, and metrics that map to user needs and design requirements.
  • Productionize models: inference performance, reproducibility, monitoring for drift/regressions, and safe fallbacks.
  • Collaborate on hybrid algorithms, incorporating physics and learned priors, denoisers, learned regularizers, and quality estimation.
  • Help build tooling for rapid experimentation as well as rigorous verification of algorithm changes.

WHAT WE’RE LOOKING FOR

  • Strong applied ML experience plus comfort with signal processing / imaging or adjacent domains.
  • Ability to move fluidly between research prototypes and production-quality systems.
  • Strong evaluation discipline: metrics, ablations, data leakage avoidance, and reproducibility.
  • A demonstrated track record of applying ML to physics-based or inverse problems (i.e., shipped projects, a portfolio, or publications.)

USEFUL EXPERIENCE

  • ML for imaging/inverse problems (or adjacent) with strong evaluation discipline and comfort with GPU performance constraints.
  • Pragmatic production mindset: reproducible training/inference, regression testing, and safe deployment in high-stakes contexts.
  • A background in computational physics or scientific computing.
  • Leverage ML-based methods such as PiNNs and Neural Operators to solve partial differential equations arising in ultrasound simulation and imaging.
  • Experience in Agentic-SciML is a plus.
  • Hands-on experience with data curation for ML: building datasets from messy, real-world sources, defining ground truth, and managing labeling or simulation pipelines.
  • Background in data assimilation: combining observations with physics-based models (Kalman filtering, variational methods, ensemble approaches, or learned variants).

Apply at Midjourney

Or fast-track your application with a Beaverhand AI interview