Cerebras Systems builds the world's largest AI chip, delivering industry-leading training and inference speeds.
Works with leading model labs, global enterprises, and AI-native startups; OpenAI announced a multi-year partnership to deploy 750 megawatts of scale.
Responsibilities
Design, implement, and maintain APIs for chat completions, text generation, streaming, model configuration, tool calling, structured outputs, and multimodal inputs.
Create consistent request and response semantics across GPU prefill, Cerebras decode, and heterogeneous inference backends.
Integrate emerging foundation models, tokenizers, prompt formats, sampling methods, attention variants, and multimodal inputs.
Maintain API compatibility, versioning, deprecation, validation, and backward compatibility practices.
Integrate custom inference services with vLLM, PyTorch, Hugging Face libraries, AMD ROCm stack, and Cerebras runtime components.
Build control and data paths for disaggregated inference: request routing, state transfer, error handling, retries, lifecycle management.
Optimize streaming, time to first token, latency, throughput, batching, serialization, tokenization, scheduling.
Build validation systems for tokenization, sampling, logits, outputs, precision changes, determinism.
Define service indicators and build logging, tracing, metrics, dashboards, health checks, diagnostic tooling.