Performance Engineer, Inference

Sarvam serves language, speech and multimodal models across a multi-node GPU fleet. You will optimize and operate distributed inference serving, train speculative models, and own system performance metrics.

• Operate and extend distributed production serving for large models
• Integrate model and kernel artifacts into multi-node serving stacks
• Train draft models and tune speculative decoding acceptance rates
• Measure latency, throughput, GPU utilization and serving costs
• Own inference service-level objectives and collaborate across engineering teams

Experience

5+ years

Domain

ai infrastructure

Sarvam has 28 open roles in Bangalore, at 732, Chinmaya Mission Hospital Road, Indiranagar Stage 1, Bengaluru, Karnataka 560038. See every Sarvam role.