Jiasheng Gu

Research Engineer · Shanghai, China · On-site

• Work on artificial intelligence and Python systems for robotic foundation models, large language models, and vision-language models for embodied agents.
• Contribute to world-action model research for generalist robot policies and cross-embodiment transfer.
• Build research systems at the intersection of multimodal foundation models, robot learning, and embodied AI.

Algorithm Engineer · Hangzhou, China · On-site

• Worked on Python-based algorithm engineering.

Algorithm Engineer · Hangzhou, Zhejiang, China · On-site

• Worked on large language model systems with PyTorch.

Honor M.S.

• GPA: 4.0/4.0
• Relevant coursework: deep learning, probability, linear algebra, parallel computing

B.E. in Telecommunications Engineering

• GPA: 3.6/4.0, Top 10%
• Relevant coursework: data structures, calculus, discrete mathematics

Research collaborator · Mentor: Wenpeng Yin

• Studied robustness of learning from task instructions.
• Experimented with and analyzed instruction-tuned models under perturbed instructions.

Research collaborator · Mentor: Dongkuan Xu and Xipeng Shen

• Worked on zero-shot code generation via rule-AI co-learning from documents.
• Proposed a framework combining rule-based and AI-based methods to generate DSL code.

Research collaborator · Mentor: Pengfei Liu

• Trained LLaMA-13B through continued pretraining and finetuning on curated math datasets.
• Improved model performance on mathematical reasoning tasks.

Research collaborator · Mentor: Peter A. Beerel

• Designed visual networks with very low FLOPs.
• Proposed and evaluated dilated depthwise convolution for capturing global information.

Research collaborator · Mentor: Soroush Vosoughi

• Analyzed the evolution of artificial intelligence techniques in biomedical publications.
• Developed an analysis method for biomedical AI publication trends.

Research collaborator · Mentor: Yuyi Wang

• Designed pre-training tasks for text summarization.
• Used learned metrics to identify high-importance sentences as summaries for more effective pre-training.

SDE Intern

• Improved system scalability by refactoring timed feature extraction and computation.
• Optimized timed computations to improve overall system efficiency.

NLP Intern

• Designed and implemented a hybrid long-text summarization system.
• Combined DGCNN-based extractive summarization with BART-based abstractive refinement.

Teaching Assistant

• EE 503: Probability for Electrical and Computer Engineers