Principal Engineer - Machine Learning

Job description

About This Role — The Mission Most ML engineering roles at large companies mean contributing to a platform team where your work disappears into a pipeline that fifty other engineers also touch. This role is different. You will be the primary owner of the ML systems that detect product development defects, model material behavior with limited data, and select the highest-value experiments from an active learning pipeline.

Your models will run in product development. Your decisions will matter immediately.

Key Responsibilities

Deep Learning Model Implementation & Product Development Ownership: Build, train, evaluate, and maintain CNN/U-Net/ViT models for automated inspection and measurement. Own model performance end-to-end — ablation studies, confidence calibration, product development performance monitoring.

Anomaly Detection Systems

Build and maintain real-time anomaly detection for product development sensor and time-series data streams — statistical baseline, threshold calibration, drift alerting. Sole implementation owner for this workstream.

Surrogate Modeling & Active Learning Operations

Own implementation and iteration of surrogate model pipelines and active learning systems under Technical lead’s architectural direction. Configure acquisition functions; integrate with versioned feature sets.

Data-to-Model Interface Ownership

Own the data contract between the Data Engineer and the ML model stack. Define feature specifications, validate datasets against model input requirements, and escalate data quality issues before they reach the training pipeline.

MLOps Maintenance & Product Development Reliability

Maintain model versions, training pipelines, and containers under platform architecture. MLflow tracking, CI/CD contribution, product development monitoring, and degradation escalation.

Junior Mentorship & Documentation

Provide code review guidance to team; document model design decisions and evaluation outcomes to production-handoff standard.

Qualifications

Requirements

Education: Bachelor's or Master's degree in AI, Machine Learning, Computer Science, or related field. AI major or strong AI research focus preferred.

Experience

1–3 years of hands-on ML engineering experience, or equivalent depth demonstrated through internships, academic research, or open-source contributions. Must show component-level technical ownership within an end-to-end ML pipeline (training through deployment) — not just execution under direction. Kaggle rankings, arXiv preprints, or significant open-source ML contributions are valued as evidence of depth.

Must have Skills

Python: Strong proficiency — primary ML development language PyTorch: Proficient → Expert — independent model training and evaluation Computer Vision: Strong foundation in CNNs, plus hands-on depth in at least one of: U-Net/segmentation, ViT/transformer-based vision, or time-series anomaly detection. Candidates with depth across multiple areas (e.g. full inspection-scope coverage — segmentation, transformer vision, and anomaly detection together) will be considered for the higher end of the band.

Surrogate Modeling

Implement and iterate surrogate pipelines under P110 architectural guidance Active Learning: Configure acquisition functions; uncertainty-guided experiment scheduling Data-to-Model Interface: Define feature specs; validate incoming datasets against model requirements; flag data quality issues before training Practical MLOps: MLflow, Docker, Git, basic CI/CD contribution Model Evaluation & Uncertainty Analysis: Ablation studies, confidence calibration, validation methodology Technical Documentation: Model design decisions and evaluation results to production-handoff standard Good to have Skills: PINNs implementation under technical guidance Bayesian methods — Bayesian neural networks, Gaussian processes, ensemble uncertainty, calibration Reinforcement learning basics — gym environments, policy gradient concepts AWS fundamentals — S3, EC2, SageMaker basics; entry-level cloud ML deployment RAG pipeline fundamentals