DI ZHANG
San Jose,CA
Summary
Innovative ML Model Development Engineer with expertise in designing and quantizing computer vision, audio, and language models for edge devices. Skilled in developing large-scale distributed training systems and implementing hardware-aware neural architecture search. Focused on enhancing machine learning applications through efficient solutions that drive high-impact projects.
Overview
8
8
years of professional experience
Work History
ML Model Development Engineer
TetraMem
01.2023 - Current
- Architected an end-to-end quantization and model optimization toolchains supporting A8W8 and A8W4 precision execution, significantly improving model performance and efficiency across multiple generations of company-specific edge AI chips.
- Deployed lightweight computer vision and audio ML models—including real-time face detection, visual wake words, and audio denoising demos—optimized for proprietary edge AI chips by executing QAT/PTQ and leveraging hardware-aware NAS.
- Compressed a TinyBERT reranking pipeline to an extra-lite footprint (< 4M weights) by applying Quantization-Aware Training (QAT) coupled with Noise-Aware Training (NAT), mitigating low-precision degradation to yield a 0.5246 MRR@10 on-chip.
- Integrated new operator support (ConvTranspose, Uint16) into a Rust-based ML compiler SDK, enabling comprehensive on-chip model execution and testing for next-generation hardware.
- Directed intern research initiatives—focusing on music classification and edge-optimized "Tiny Stories" models—by defining technical scopes, architecting project roadmaps, and guiding end-to-end execution.
Algorithm Engineer
Aibee Inc.
08.2019 - 10.2022
- Delivered multi-server parallel deep learning training systems (data + model parallel), reducing face recognition training time from 4 weeks to 1 week.
- Implemented automated evaluation pipelines for AI product release at scale using Airflow, Kubernetes, and Argoflow with Streamlit dashboards.
- Designed CI/CD pipelines using GitLab CI/Jenkins to streamline AI production processes for cloud and embedded environments.
- Developed Kafka-based services to automate AI protobuf data dumping and uploading, enhancing data management efficiency.
Software Engineer for HMI
Futurewei Technologies, Inc.
08.2018 - 06.2019
- Developed and optimized Driver Status Monitoring ML applications for mobile platforms, enhancing face detection and gaze estimation using MobileNet and YOLO models.
- Deployed algorithms on Linux and Android (Futurewei NPU) via JNI and C++ APIs, ensuring seamless integration and performance.
Education
Master of Engineering - Electronic Computer System Engineering
Rensselaer Polytechnic Institute
Troy, NY01-2018
Bachelor of Engineering - Information Engineering
Shanghai Jiao Tong University
Shanghai, China01-2016
Skills
- Programming Languages: Python, C, Rust, Java, SQL
- Technical Frameworks & Tools: PyTorch, TensorFlow, OpenCV, ONNX, NNI, Kubernetes, Airflow, Argoflow, Flask, Git
Timeline
ML Model Development Engineer
TetraMem
01.2023 - Current
Algorithm Engineer
Aibee Inc.
08.2019 - 10.2022
Software Engineer for HMI
Futurewei Technologies, Inc.
08.2018 - 06.2019
Bachelor of Engineering - Information Engineering
Shanghai Jiao Tong University
Master of Engineering - Electronic Computer System Engineering
Rensselaer Polytechnic Institute