2027届校招-算法和解决方案工程师
北京、杭州、上海
校招
正式
研发 - 电子 / 半导体
职位 ID:A227994
职位描述
1、跟进前沿算法技术、开源框架及训推并行加速技术,协同客户或上游团队,完成算法选型或迭代。 2、协同系统架构等团队,输出算法部署和对应的系统架构方案, 完成方案原型搭建。 3、协同工具链等团队,解决算法部署过程中的精度、速度、存储限制等问题,推进项目落地。
职位要求
1、编程基础扎实:熟练掌握Python、C、C++等主流开发语言,具备扎实的代码编写、调试及工程开发能力,可独立完成算法代码开发与优化。 2、熟悉深度学习基础理论、训练策略及模型优化方法,掌握PyTorch等主流深度学习开源框架,理解算法量化,量化感知训练等量化方案。 3、深入理解神经网络基本工作原理,掌握CNN、LSTM/GRU、Transformer+Attention、MoE等核心网络架构的原理、特性及适用场景,可根据业务需求完成模型选型与替换。 4、具备极强的自主学习能力和环境适应能力,能够快速跟进前沿算法技术与工程方案;拥有优秀的逻辑分析、问题排查和拆解能力,可高效解决算法研发与落地过程中的各类技术问题。 5、具备优秀的跨团队沟通能力和团队协作意识,责任心强、执行力突出,能够高效推进项目落地,抗压能力良好。 加分项: 1、具备大语言模型、生成扩散模型、多模态大模型、图像、视频、语音、TTS等至少一个及以上方向的算法研发与工程落地实战经验,可独立完成算法从训练到上线的全流程落地。 2、推理加速与工程优化能力:熟悉TVM、ONNX、TensorRT、NCNN、MNN,vLLM,llama.cpp等主流深度学习推理框架及加速平台,掌握模型量化、算子优化、推理链路精简等工程优化手段;具备大模型KV Cache优化、推理加速、显存及算力资源优化、模型轻量化等相关经验。
投递

Founded in 2017, Witmem focuses on the field of computing-in-memory chips, innovatively uses Flash memory to complete the computing-in-memory of neural networks, solves the problem of AI storage walls, improves computing efficiency and reduces costs.
The R&D team is formed by Dr. Wang Shaodi and Dr. Guo Xinjie in conjunction with a number of scholars and industry practitioners. They have an average of more than 10 years of industrial work experience.
Witmem's WTM2101 chip is suitable for low-power AIoT applications, and can complete large-scale deep learning operations with microwatt to milliwatt power consumption, especially suitable for intelligent voice and intelligent health services in wearable devices, and has completed mass production and market applications . WTM8 series chips are aimed at 6-48Tops computing power products and are used for real-time processing of 4K-8K video.
In January 2023,Witmem announced the completion of the B2 round of financing. In the future, Witmem will continue focusing on the field of computing-in-memory chips and lead the industrialization of computing-in-memory.
Website: https://en.witmem.com/
YouTube: https://www.youtube.com/@Witmem2017
Headquartered in Beijing, China.