Haoming Guo

Henan Changlv Group Real Estate Co,. Ltd.

• Implemented an AI-driven system for parking management, encompassing occupancy tracking and multi-violation detection.
• Implemented a PyTorch dataset class for KITTI and built the KITTI API by transforming the label to coco-like format.
• Fine-tuned a pre-trained DETR model by customizing its final classification layer to align with KITTI’s label mapping and conducting extensive hyperparameter optimization, resulting in a substantial increase (12%) in detection performance as measured by average AP and enabling seamless model evaluation with enhanced performance.

USC INI (Mark and Mary Stevens Neuroimaging and Informatics Institute)

• Developed a DCN-GAN in PyTorch for cortex thickness mesh generation, enhancing data augmentation.
• Designed model with ResNet blocks and attention layers, reducing FID score by 41.52%.
• Optimized model on GPU clusters using Omegaconf and implemented SciPy mesh visualization.
• Improved GAN stability using WGAN and WGANGP techniques.

Eth Tech

• Implemented distributed e-commerce MongoDB system and backend CRUD operations with Tkinter/PyMongo.
• Developed a Streamlit-based query interface.
• Designed MongoDB sharding key, reducing query latency by 50%.
• Built a real-time DB on AWS EC2 using REST API/cURL.
• Visualized transaction data using D3.js, Vue, and Deck.gl.

Yisen Tech

• Developed a hybrid recommendation system using Yelp data.
• Processed large-scale data using Spark RDD.
• Improved collaborative filtering with an XGBoost model, achieving 95.1% accuracy (6% boost).