大型模型数据算法工程师（多模态/感知/仿真方向）

<p>Hong Kong flavor responsibilities:</p><p>1. Data processing and optimization: Responsible for the data acquisition, cleaning, annotation, and optimization of the embodied intelligence large model, and building a data-model closed loop.</p><p>2. Algorithm development and deployment:</p><p>a. Multimodal direction: Use multimodal models to process text, image, and video data, and provide key data processing components for the data management platform.</p><p>B. Perception direction: optimize the end-to-end data transmission link (audio and video encoding and low-latency transmission), and deploy multimodal algorithms.</p><p>C. Simulation direction: generate high-fidelity simulation data with physical engines (Mujoco/isaacim), design data augmentation and reinforcement learning strategies.</p><p>3. Tools and system construction: Develop a data full chain tool to improve data processing efficiency; optimize computing performance at the edge or in a simulation environment.</p><p></p><p>Qualifications:</p><p>1. Basic skills:</p><p>a. Proficient in Python, familiar with Linux/Git, solid foundation in data structures and algorithms, good coding conventions and engineering skills. </p><p>b. Understand CV/NLP/multimodal models, with model optimization (quantization, pruning) or big data processing (Spark/Flink) experience being even better.</p><p>c. Computer Science and Technology, Data Science, Information Engineering, Electronic Information and other related professional graduation. Bachelor's degree or above.</p><p>2. Directional requirements (satisfy one of the following):</p><p>a. General direction: Familiar with large models or multimodal algorithms, able to independently train models and develop data processing tools. Data can be processed in large-scale using distributed technology.</p><p>b. Real machine direction: Familiar with network transmission protocols (UDP/TCP), audio and video coding (H264/H265/AAC, etc.), sensor integration and edge deployment (TensorRT/Jetson).</p><p>c. Perception direction: Have physics engine (Mujoco/isaacim) or synthetic data generation (GAN/RL) experience. Understand reinforcement learning, imitation learning or synthetic data generation techniques (such as GAN, Diffusion Models)</p><p></p><p>a. Able to independently research new technologies, solve practical problems, and have good team spirit. </p><p>b. Have python front-end and back-end development skills or hardware debugging experience.</p><p>c. Computer/Electronics/Robotics-related bachelor's degree, those with papers, open-source project contributions or large factory experience are preferred.</p>