混元多模态大模型算法加速工程师（深圳/北京/上海/杭州）
Core
Accelerating multimodal large models (image/video generation, understanding, speech) via algorithmic optimization and inference efficiency improvements.
Role type
Senior IC algorithm acceleration engineer (multimodal LLMs)
Builds
Optimized inference pipelines for image/video generation, audio understanding/generation, and multimodal unified modeling
Domain
Generative AI, Multimodal Learning, Model Compression
Deliverable
production ML models
Required skills
Python, PyTorch, Transformer/Diffusion architecture expertise, model quantization, pruning, distillation, speculative decoding, KV Cache optimization, DPO/RLHF alignment
Preferred skills
Experience with attention sparsity, real-time dialogue systems, visual multi-turn interaction
Technologies
PyTorch, Diffusion models, Transformer, ASR, TTS
Responsibilities
Track academic/industrial innovations in generative and multimodal models; Design algorithmic optimization schemes to reduce end-to-end latency; Bridge algorithm and framework teams to deploy solutions for image/video/audio tasks; Collaborate with business teams to ensure technical implementation.
Seniority
Senior, hands-on IC