Junior ML Engineer — Customer Support Email Classification

We're a SaaS company. 300+ support emails per day. Right now a person reads each one and routes it to the right queue: billing, technical, account management, feature requests, or complaints. We want a classification model to do this automatically. We have 40k labelled historical emails. We want 90%+ accuracy on the top 5 categories, with a fallback to human review for low-confidence cases. We also need it to extract key fields: customer tier, urgency, sentiment, and whether there's an SLA breach risk. Simple project. We want it done right.