Senior Machine Learning Engineer, Classification and Detection

About the job

This is a role for someone who trains models. Not someone who calls them. The company is building classification and detection systems: fine-tuned transformers, entity detection over structured and unstructured content, and behavioral scoring. Production model ownership required - from data through training through evaluation through the retraining loop.

About SmartVerify

Building the data egress control plane for enterprise AI, sitting inline between AI agents and enterprise data to inspect queries, enforce policy in real time, and produce audit trails. Greenfield build with freedom to update specifications.

What You Would Own

• Behavioral classification model: multi-class classification of AI agent query intent, producing labels, confidence, and evidence
• PII and PHI detection over query content and returned data with span-level identification
• Behavioral risk scoring combining deterministic and model-derived signals
• Model training, serving, and versioning on SageMaker
• Evaluation harness, drift detection, and retraining loop with co-op support
• Classification output contract for downstream consumers

What We Are Looking For

Must have: Production ML ownership; Transformer fine-tuning for text classification; Sequence labelling or named entity recognition; Python, PyTorch, Hugging Face Transformers; Evaluation rigor; Comfort with written design specs; Authorization to work in Canada or US, located in BC or Seattle area

Nice to have: Weak supervision or LLM-assisted labelling; Anomaly detection; SageMaker experience; Regulated domain experience (HIPAA, PCI DSS, GDPR, SOC 2); Fraud/abuse/security detection background; Distillation or quantization; Kinesis/Kafka; SQL and query parsing; Mentoring experience

Who Does Well Here

You've shipped production models others depended on and remember what broke. Honest about model limitations. Can read architecture documents and respectfully disagree with reasoning. Comfortable being the only person understanding this layer and document accordingly. Prefer ownership over large teams.

The Stack

Intelligence layer: PyTorch, Hugging Face Transformers, SageMaker training and inference, evaluation and drift tooling. Infrastructure: AWS, EKS, Terraform, Helm, Prometheus, Grafana.