aiInternal Project
Product Recommendation System
Image-based product search and recommendation system with 92% success rate.
Developed at Huawei
Senior Research Engineer
2021-2022
Internal project - no public access
Screenshots
About This Project
Developed an advanced product recommendation system that uses computer vision and deep learning to provide personalized recommendations. The system processes product images to understand visual similarities and combines this with user behavior data to deliver highly relevant suggestions. Also implemented ad click prediction using Deep & Cross Networks.
Key Features
- Image-based product similarity search
- Personalized recommendation engine
- Deep & Cross Network for ad click prediction
- FTRL optimization for CTR prediction
- Scalable processing with Hadoop and Spark
- Real-time recommendation serving
Challenges & Solutions
- Processing millions of product images efficiently
- Balancing recommendation accuracy with diversity
- Scaling to handle high-traffic production loads
- Cold-start problem for new users and products
Results & Impact
- 92% success rate in product recommendations
- 90% accuracy in ad click prediction
- Streamlined CI/CD with MLOps platform
Technologies Used
PythonPyTorchSparkHadoopComputer VisionDCNFTRL
Project Details
- Category
- ai
- My Role
- Senior Research Engineer
- Duration
- 10 months
- Year
- 2021-2022
- Company
- Huawei
- Status
- Internal / Private
Internal Project
This project was developed for internal use at Huawei. Source code and live demo are not publicly available due to confidentiality.