Summary
What you’ll impact
The Senior Machine Learning Engineer will design and own an anomaly‑detection system for ultrasound‑based battery cell inspection, ensuring it scales from gigafactory production to lab environments. This hands‑on role combines deep learning model development with end‑to‑end ML pipeline ownership, deployment, and traceability.
Responsibilities
What you'll do
- Design, build, and iterate the anomaly-detection approach that flags abnormal cell scans, and keep it current with the state of the art.
- Think structurally about running it at production scale: throughput, traceability, model versioning, reproducible and standardized evaluation, and drift as new sites and datasets come online.
- Own the ML pipeline end to end as production software: containerized training and inference, orchestration, CI/CD for models, experiment tracking and model registry (we use MLflow), and lineage across data, code, and artifacts.
- Build out Titan’s Battery Quality Library, including a scalable annotation workflow and extraction of multi-modal data (electrical, teardown, CT) for training and validation.
- Define quantifiable anomaly metrics fit for production use and grounded in the physics of the cell, such as its compositional structure and how ultrasound propagates through it.
- Package models to run wherever the product runs, in the cloud and on-prem at the edge, and monitor them once they are out there.
- Support new and existing pilots across Europe, Asia, and North America, informing implementation strategy with what works for gigafactory inbound inspection.
- Document methods, key algorithms, and their evaluation clearly, and prepare the material that drives fast, evidence-based stakeholder decisions.
- Set technical direction for the ML work, raise the bar through code review, and mentor engineers as the team grows.
- Partner with Product, Battery Science, Firmware, Software, and Data Science teams to advance and improve existing software and develop dependable production capability. Software Engineering owns the platform, edge runtime, and application; you own the models and the pipelines that produce and serve them.
Requirements
What you’ll bring
- Strong applied machine learning background designing and training deep learning models (PyTorch, Tensorflow, or JAX).
- Built generative anomaly detection models, such as diffusion models, and familiar in detail with their training regime and tuning of the noise/reconstruction schedule.
- Familiarity with a range of approaches to AD, such as prototype-based methods, localization, segmentation, and working with anomaly maps, and the judgment to select among them rather than defaulting to one.
- Owned model evaluation end-to-end: dataset construction, imbalance-aware metrics, and calibrated operating points.
- Owned reproducible ML pipelines with lineage across artifacts, best practices for reproducible experiments.
- Strong software engineering fundamentals in Python: testing, packaging, code review, and building services other teams depend on.
- Docker/container workflows and CI/CD ownership for training and inference.
- Production MLOps: remote model deployments, drift/regression monitoring, and the tooling to catch problems before customers do. We run training and inference on AWS, including SageMaker.
- Experiment tracking and a model registry used in earnest, such as MLflow: runs, metrics, artifacts, and a clear path for promoting a model to production.
- Self-sufficient across a modern micro-service stack, with production AWS experience.
- Senior enough to make the call on approach, defend it with evidence, and carry it through to something running in production (typically 7+ years of relevant experience).