Email Classification (spam or not) Service
Design and deliver a highly accurate, cost-efficient, and enterprise-grade email classification platform that determines whether an incoming email should be classified as SPAM or NOT-SPAM for Outlook-like workloads. The system must leverage a Mistral-based ML model, integrate seamlessly with existing email ingestion pipelines, and operate at global scale while maintaining customer trust, security posture, and operational excellence.
Functional Requirements (FR)
Classify each incoming email as:
- SPAM
- NOT_SPAM
- Provide low-latency inference suitable for near-real-time inbox delivery.
- Support model versioning and safe rollout (blue/green or canary).
- Capture classification confidence and decision metadata for auditability.
Non-Functional Requirements (NFR)
- Scale & Performance
- Throughput: ≥ 100 emails/hour (baseline)
- Latency: P95 ≤ 200 ms per email classification
- Concurrency: Elastic scaling with burst tolerance (Bias for Action)
- Availability & Reliability
- SLA: ≥ 99.9% availability
Design and deliver a highly accurate, cost-efficient, and enterprise-grade email classification platform that determines whether an incoming email should be classified as SPAM or NOT-SPAM for Outlook-like workloads. The system must leverage a Mistral-based ML model, integrate seamlessly with existing email ingestion pipelines, and operate at global scale while maintaining customer trust, security posture, and…