搜狗输入法-打字引擎推理优化工程师
Core
Develop and optimize the typing engine kernel for Sogou Input Method, focusing on architecture upgrades, core component R&D, and adapting large models for various hardware platforms to improve accuracy and response speed.
Role type
Senior IC machine-learning engineer (input method optimization)
Builds
End-to-end typing engine architecture and optimized large model inference components for input methods
Domain
Consumer software / Large model inference optimization
Deliverable
production ML models
Required skills
C/C++, data structures and algorithms, large model inference frameworks (vLLM, SGLang, TensorRT-LLM), attention mechanisms, KV Cache management, Batching, Prefill/Decode optimization, quantization, operator fusion, memory management
Preferred skills
Complex problem solving, independent development and debugging, industry trend tracking
Technologies
vLLM, SGLang, TensorRT-LLM
Responsibilities
Upgrade end/cloud typing engine architecture; Develop and improve core input logic; Adapt large models to different hardware platforms; Explore and apply new technologies to input scenarios
Seniority
Senior, hands-on IC
