I work on neural data analysis and robot learning systems: decoding visual cortex activity at scale, and training vision-language-action policies from multimodal robot demonstrations.

Honors thesis

Functional decoding and neural tuning dynamics in visual cortex
Brandeis University, 2024–2026
Advisor: Stephen D. Van Hooser (lab)

I wrote and defended a senior honors thesis on how visual cortical neurons encode and change their tuning. Alongside the thesis I built the lab software that made the analysis possible: Python/MATLAB pipelines that turn terabytes of 2-photon calcium imaging into memory-mapped binaries, plus 50+ feature-extraction scripts (about 40% faster processing). I also maintained CI for open-source lab tools (vhlab-toolbox, NDI-matlab).

Status: manuscript in preparation.

Vision-language-action and manipulation

Visiting student researcher
Shanghai Jiao Tong University, June–Sept 2026
Advisor: Yutong Ban

I engineered a multimodal data pipeline that synchronizes dual-arm joint states with 3-channel RGB at 30 FPS (35% lower I/O latency across 100+ episodes), scaled SFT and advantage-conditioned training for an InternVL-1B VLA model on 8×A100 GPUs with PyTorch FSDP, and improved cross-task policy success by 18%. I also built a human-in-the-loop teleoperation stack for closed-loop evaluation.

Publications

• Wan, Z. Honors thesis on functional decoding and neural tuning dynamics in visual cortex. Brandeis University, 2026. Manuscript in preparation.

Portfolio · CV · Download PDF