CV

Graduate Research Assistant

Automated Pedicle Screw Trajectory Planning

• Advisor: Prof. Mathias Unberath, Ph.D. Blanca Inigo Romillo
• Developed a Monte Carlo-based path planning pipeline for pedicle screw placement across the full thoracolumbar spine (T1–L5), processing 742 patient CT-derived mesh models and generating 23,000+ validated trajectories with 99% success rate.
• Engineered region-adaptive constraints for 4 anatomical zones to address varying pedicle morphology, and implemented a surface normal-based endplate filter that eliminated 100% of superior/inferior surface violations.
• Built automated batch processing infrastructure with mesh quality control, EDT-based clearance computation, and FCSV output for 3D Slicer visualization and validation.

Conditional diffusion model for surgical view planning under C-arm guidance

Action-Conditioned Intraoperative X-ray View Prediction

• Built an observation prediction module for closed-loop surgical view planning: given a current fluoroscopy image and a candidate 6-DoF C-arm motion, a conditional diffusion model (DDPM training / DDIM inference) predicts the X-ray at the resulting viewpoint as a visual preview for a downstream VLM (MedGemma).
• Constructed a DRR training dataset from 827 CT volumes using DeepDRR, sampling 100 poses per case (5 vertebra centers × 20 angles spanning AP/lateral/oblique) and filtering ~1,500 training pairs per case by angular and translation distance thresholds.
• Designed a U-Net backbone with cross-attention for source image conditioning and AdaGroupNorm for injecting a 9-D relative pose embedding (6D rotation + 3D translation).
• Trained with DDPM training / DDIM inference using mixed-precision (fp16) on an RTX 3090.

Intern at Software Development Department

Qingdao Haier Technology Co., Ltd

• Utilized Redis to optimize the system performance bottlenecks under high concurrency; reduced the database visits by approximately 40% and minimized the overall response time.
• Designed and developed a batch-processing mechanism to minimize memory consumption.
• Optimized database query statements, avoided complex nested queries and full table scanning, and added indexes to improve database query performance in high-concurrency scenarios.
• Implemented a delayed double-deletion strategy, as opposed to the asynchronous listening and reliable message-deletion strategy.

Intern at Software Development Department

Qingdao Haier Technology Co., Ltd

• Ensured high availability of the user experience cloud platform and resolved performance crashes; designed the app using the SpringBoot and utilized PostgreSQL for data storage and table creation.
• Used Profiler to specify code and performance bottlenecks; implemented strategies such as algorithm optimization and reduction of database access frequency.
• Leveraged Alibaba Cloud’s Prometheus service to evaluate memory utilization patterns, specifying the primary causes such as excessive object creation, memory leaks, and large object presence.
• Adopted the Object Pool Technology to create common objects beforehand, batch-processed big data to reduce memory utilization, and used Bitmap to compress user information data structures.

Intern at Software Development Department

iFLYTEK Co., Ltd.

• Developed a domain-specific speech recognition pipeline for hydrology and water conservancy engineering.
• Collected and processed fieldwork data, extracting key clauses and terms from technical specifications through custom text-segmentation scripts.
• Built a speech synthesis corpus using Tacotron2_CSMSC, generating multi-timbre audio and aligning text–audio pairs into training, validation, and test datasets.
• Trained and evaluated multiple ASR architectures—DeepSpeech2, Transformer, Conformer, and iFlytek’s open-platform model—on the domain corpus.
• Selected 10–25 top-performing checkpoints based on validation loss and combined them into an optimal averaged model.
• Achieved a 30% improvement in recognition accuracy, reaching a >99% success rate on domain-specific evaluation data.

M.S.E

Johns Hopkins University

Computer Science

• Department of Computer Science, Whiting School of Engineering

B.Eng.

Shanghai Institute of Technology

Software Engineering

• CGPA: 4.0/5.0 (Average Score: 90%; Rank: 1/408)

National Scholarship

Shanghai Institute of Technology

Outstanding Student

Shanghai Institute of Technology

University-Level First-class Scholarship

Shanghai Institute of Technology