Xariv Engineering
Case studies & engineering deep-dives from the Xariv team
AI Infrastructure
Networking Bottlenecks in Large-Scale Mixture-of-Experts (MoE) Inference
Architecture Study. This is a hypothetical system study. The numbers, topology, and traffic patterns are illustrative and do not descr...
Architecture
Designing an Idempotent Webhook Pipeline
Every webhook provider retries on failure, which means your handler will see the same event twice. If processing isn’t idempotent, you ge...
Performance
Cutting P99 Latency by 60% on Our Search Service
When our search API started missing its latency SLO, the averages looked fine — it was the tail that hurt. This is a walkthrough of how w...