大模型推理研发实习生（J106486）
Core
Develop and optimize large model inference systems, focusing on deployment, performance tuning, and engineering capabilities for online and offline inference services.
Role type
Intern, Large Model Inference R&D
Builds
High-performance inference engines, service frameworks, and deployment pipelines for large language models
Domain
Artificial Intelligence / Large Language Models / System Optimization
Deliverable
production ML models
Required skills
Python, C/C++, Go, Java, Linux, Model Compression, Quantization, Operator Optimization, KV Cache Management, Batch Scheduling, Multi-card/Multi-machine Inference, Distributed Inference (TP/DP/PP/EP), Transformer Architecture, GPU Memory Management
Preferred skills
Experience with Inference Engines, Model Deployment, Service Frameworks, Performance Optimization, Heterogeneous Hardware, Paged Attention, Dynamic Batching, Observability Systems
Technologies
Python, C, C++, Go, Java, Linux, GPU, Transformer, KV Cache, TP, DP, PP, EP
Responsibilities
Develop and optimize inference systems and service frameworks; Optimize inference performance including latency, throughput, and memory utilization; Handle production environment issues like performance jitter and memory bottlenecks; Build backend capabilities for model service management and observability; Analyze and resolve complex system performance and stability issues.
Seniority
Intern
