juin 2023 - Présent
Primary responsibilities include implementing and optimizing AI algorithms for audio and image recognition, focusing on IP Camera systems. Developed real-time video inference code. Worked on hardware platforms, including Ambarella and Hailo (AP, DSP, NPU), ensuring efficiency in constrained environments. Ensured reliable operation in real-world applications like surveillance and security systems.
Key Achievements and Responsibilities:
1. Audio AI Model Development
- Designed and implemented models for event detection, noise reduction, and echo cancellation.
- Integrated inference code into edge devices, ensuring real-time performance in diverse environments.
2. Full-Stack Development
- Proficient in Python, C/C++, with complete inference pipelines.
- Applied signal processing techniques like Fourier Transform, windowing, and MFCC.
3. Cross-Platform Optimization
- Optimized models on Intel, ARM CPUs, and NPUs like Ambarella and Hailo.
- Familiar with Nvidia GPU optimization.
4. Deep Learning Expertise
- Mastered RNN, LSTM, GRU, and Transformer.
- Skilled in PyTorch, TensorFlow, and ONNX.
- Researched quantization techniques and explainable AI.
- Applied knowledge distillation to compress models while maintaining performance.
Technical Skills: Deep Learning | Computer Vision | Audio Processing
Signal Processing | Multi-Platform Development | Algorithm Optimization
Python | C/C++ | PyTorch | TensorFlow | ONNX