Qifei Cui

Qifei Cui

he/him/his

My research seeks intelligence that flows naturally through constraints — whether teaching machines to move with human grace or to reason with mathematical clarity.

Sitemap

A list of all the posts and pages found on the site. For you robots out there, there is an XML version available for digesting as well.

Pages

Posts

Black Truffle Basque Cheesecake: A Fusion of French Elegance and Spanish Tradition

3 minute read

Published:

The Basque cheesecake has taken the culinary world by storm with its signature burnt top and creamy interior. This recipe elevates the classic by incorporating luxurious black truffle and aged Comté cheese, creating a sophisticated dessert that bridges the gap between sweet and savory. We present Black Truffle Basque Cheesecake—a decadent fusion perfect for special occasions.

GRPO-RS: A Next-Gen Graph-based Reinforcement Learning Framework for Recommender Systems

5 minute read

Published:

Recommender systems (RS) have seen rapid evolution in recent years, especially with the rise of Graph Neural Networks (GNNs) and Reinforcement Learning (RL). However, most existing solutions still struggle with challenges such as data sparsity, cold start, and the long-tail effect. We introduce GRPO-RS—a novel architecture combining graph signal regularization and reinforcement learning for offline and scalable recommendation optimization.

Bridging HPSTM and SKEL

1 minute read

Published:

Feeding 46 biomechanical angles into the kinematic layer is the moment HPSTM truly shakes hands with SKEL.

Camera Pose Expressions for HPSTM-Skeleton-NeRF Integration

2 minute read

Published:

Skeleton serves as an explicit warp between canonical space and every posed frame. The render-and-compare loss feeds back dense gradients that refine both the NeRF and the HPSTM joint angles simultaneously.

From HPSTM to HPSTM-Gen

2 minute read

Published:

Why we’re turning our pose-smoothing Transformer into a generative motion model for VLA research

Tired of Coding? Try This Effortless Korean Short Rib Bake!

2 minute read

Published:

After a long day wrestling with code and debugging errors, sometimes the best therapy is a break away from the screen—preferably, into the kitchen. If you’re looking for something rewarding but not too complicated, this Korean-style baked short rib stew is the perfect dish to revive both your mood and your energy.

Pride 2025: Be an Ally by Raising Up a Flag

1 minute read

Published:

Happy Pride Month! June is a time for celebration, reflection, and showing support for the LGBTQ+ community. As developers, we have a unique canvas for expression: our own websites. Adding a small touch of Pride to a personal site is a wonderful way to signal that it’s a safe, inclusive space and to stand as an ally.

experience

projects

PDE Solvers Collection

Published:

Numerical methods for solving partial differential equations with applications in physics and engineering

NLP Claims Analysis

Published:

Natural language processing system for automated insurance claims analysis and classification

Anime Recommendation System

Published:

Content-based and collaborative filtering recommendation system for anime using user ratings and metadata

Deep Interest Network for CTR

Published:

Implementation of Deep Interest Network for click-through rate prediction in recommendation systems

Credit Risk Modeling

Published:

Advanced credit risk assessment using machine learning techniques for loan default prediction

XGBoost Fraud Detection System

Published:

Machine learning system for credit card fraud detection using XGBoost with advanced feature engineering

Convex Optimization Toolkit

Published:

Implementation of modern convex optimization algorithms with applications in machine learning and signal processing

SARIMA Time Series Toolkit

Published:

Seasonal ARIMA modeling toolkit for time series forecasting with automated parameter selection

GNN-based Recommendation System

Published:

Graph Neural Network approach to solving cold-start problems in recommendation systems with dynamic embeddings

Brain MRI Segmentation

Published:

Deep learning-based brain tumor segmentation using GANet-Seg architecture for medical image analysis

Environmental AQI Analysis

Published:

Air Quality Index prediction and analysis using machine learning and environmental data

Robotics Algorithms Toolkit

Published:

Collection of fundamental robotics algorithms including path planning, SLAM, and control

Robotics Perception and Control

Published:

Integrated perception and control system for robotic manipulation using computer vision and reinforcement learning

Health Tracker Web Application

Published:

Comprehensive web application for tracking blood glucose, ketones, food intake, and body metrics with automatic health calculations

recent

HPSTM-Gen: Generative Human Pose Sequence Model

Published:

We present a generative extension of HPSTM, enabling structured human motion to be sampled from noise using flow matching and anatomical constraints. Extensions to broader action spaces and physical robot deployment are left for future work.

research

GANNet-Seg: Adversarial Learning for Brain Tumor Segmentation with Hybrid Generative Models

Work in progress. View on arXiv

Leveraging pre-trained GANs and U‑Net, this framework combines a global anomaly detection module with iterative mask refinement under adversarial loss to accurately segment brain tumors on multi‑modal MRI. By incorporating synthetic image augmentation, it overcomes limited annotated data and achieves superior lesion‑wise Dice and HD95 performance on the BraTS benchmark, reducing reliance on fully labeled datasets for real‑world clinical use.

GitHub Repository

Genuine-ESFP: Estimating, Smoothing, Filtering, and Pose‑Mapping

Work in progress. View on arXiv

Leveraging a monocular RGB stream and a novel Transformer-based smoothing module (HPSTM), this end-to-end ESFP pipeline estimates 3D human pose via ROMP, enforces anatomical consistency through forward‑kinematics, and maps refined trajectories to a low‑cost 4‑DoF uArm in real time. By jointly predicting per‑joint uncertainty and applying dynamic filtering, it delivers smooth, reliable vision‑to‑robot imitation for desktop manipulation tasks.

GitHub Repository | Demo

HPSTM-Gen: Generative Human Pose Sequence Model

Currently ongoing research (not yet published)

We present a generative extension of HPSTM, enabling structured human motion to be sampled from noise using flow matching and anatomical constraints. Extensions to broader action spaces and physical robot deployment are left for future work.

teaching