In the lab
I'm Jingyi Zou (she/her), a researcher at Carnegie Mellon University's School of Computer Science working on tactile sensing and robot learning — building fingertip sensors, contact-rich manipulation datasets, and policies that use touch, from real hardware to simulation. I lead a physical AI team spanning two CMU research groups, the labs of Prof. Xu Zhang and Prof. Min Xu. My Ph.D. work built atomically thin synaptic devices for neuromorphic computing; now I'm teaching robots to feel.
LeRobot tactile camera plugin released.
lerobot-camera-paxini exposes Paxini PX-6AX GEN3 tactile
sensors as LeRobot cameras, covering every variant from 9 to 239
taxels. Install it from
PyPI
or read the source on
GitHub.
Four-camera tactile rig calibrated. The rebuilt SO-101 bench records a wrist camera and an egocentric view that match pi0.5 pretraining, plus two side cameras facing each other across the table for 3D reconstruction. All fixed cameras are calibrated to one shared table frame.
Tactile policy checkpoints released. Four ACT variants trained on SoTac with different tactile encoders (MLP, predictive, echo state network, and the last two combined), plus pi0.5 full and LoRA fine-tuned baselines, openly available on Hugging Face.
Tactile dataset visualizer online. A LeRobot dataset viewer extended for touch shows per-taxel force arrows, a contact timeline and the raw high-rate tactile stream next to the camera views. Try it on Hugging Face Spaces.
SoTac grows to 163 curated episodes. A 100-episode contributed batch is merged into the tactile corpus, which now spans 55,457 frames across four tasks with reproducible curation from the append-only raw set, openly available on Hugging Face.
XLeRobot fully assembled. The dual-arm mobile manipulator is standing: two SO-101 arms with fin-ray grippers, a camera mast and a mecanum wheel base on a lab cart. Wheel-mount and onboard-compute upgrades are next.
Reservoir computing manuscript in preparation. First-author paper on mechanism-anchored modelling of MoS2 synaptic devices for physical reservoir computing — bridging my device work and learning systems.
AmazingHand built and tested. The 3D-printed dexterous hand is assembled and running gesture demos through a bus servo adapter, joining the lab's dexterous-manipulation platform.
Ph.D., Electrical and Computer Engineering (2025).
Advisor: Prof. Xu Zhang. Dissertation on atomically thin 2D synaptic
devices for power-efficient neuromorphic computing.
B.S., Physics — Condensed Matter Physics (2019).
MBE growth and characterization of monolayer TMD alloys in the MBE
Lab with Prof. Yi Zhang.
Spintronics at UCLA's Device Research Lab (CSST program, Prof. Kang L. Wang), X-ray optics at the Institute of High Energy Physics, and two semesters as TA for CMU's Memory Devices and Micro/Nano Systems Fabrication courses. Cleanroom-trained hands turn out to be good at building grippers.
SO-101 gripper jaws redesigned around commercial Hall-effect tactile sensors: a magnet-embedded TPU fingertip for the MLX90393 single-taxel sensor and dual 52-taxel Paxini fingertips, with real-time recording integrated into the LeRobot stack.
A contact-rich manipulation dataset pairing single-taxel magnetic tactile sensing with wrist and top RGB and proprioception — 70 episodes of grasping and pick-and-place, openly released.
Dataset ↗Measuring how touch improves manipulation policies by climbing a ladder of tactile integrations in ACT and Pi0.5 — from baseline to tactile-as-state to learned encoders — in a policy-agnostic fork of LeRobot.
Code ↗An end-to-end imitation-learning pipeline: Isaac Sim tasks and demos, GR00T fine-tuning on a cloud GPU, and deployment to a real SO-101 through a Raspberry Pi robot client — now working on closing the sim-to-real gap.
A low-cost dual-arm, mobile SO-101 platform with fin-ray grippers and VR teleoperation, built for bimanual, contact-rich data collection with tactile integration.
A custom fingertip PCB with sixteen taxels and a microphone channel sampling at ~130 Hz, designed for dense contact-event detection on low-cost grippers.
Mobile manipulators on battery power can't afford power-hungry tactile arrays — a constraint I hit first-hand on XLeRobot. Building on my low-power device background, I'm exploring high-frequency tactile sensing that embedded robots can actually carry.
My dissertation built atomically thin MoS2 synaptic transistors that store analog weights through Maxwell‑Wagner interfacial charging in binary oxides — reaching a ~105 analog switching ratio at about 50 pW programming power, orders of magnitude below biological synapses. A physics-based compact model captures the charging dynamics and transport. (First-author paper in 2D Materials, 2024.)
One synaptic transistor's intrinsic dynamics can compute: used as a time-multiplexed physical reservoir with only a linear readout, it reaches ~93% on MNIST — and transfers to temporal signals, improving ECG arrhythmia detection by 13 points and industrial fault detection by 17.5 points over no-reservoir baselines. First-author manuscript under review.
Before devices came the atoms: MBE growth of monolayer TMD alloys with RHEED, STM and ARPES characterization at Nanjing University, then a full cleanroom pipeline at CMU — deposition, lithography, etch, measurement. Along the way I contributed to collaborative work on reconfigurable thermal metamaterial pixel arrays (Nature Communications, Nano Letters) and high-mobility tellurium photodetectors.
Mechanism-anchored modelling of MoS2 synaptic devices for efficient physical reservoir computing first author under review
J. Zou, H. Xie, N. Zhang, Y. Zhong, T. Huang, S. Lin, et al. — under review (2026)
Tunable pixelated graphene based thermal infrared emitter array and its electrothermal characterization
Y. Zhong, X. Liu, Z. Wang, J. Zou, T. Huang, S. Lin, et al. — Journal of Quantitative Spectroscopy and Radiative Transfer, 109817 (2026)
Electrically programmable pixelated coherent mid-infrared thermal emission
X. Liu, Y. Zhong, Z. Wang, T. Huang, S. Lin, J. Zou, et al. — Nature Communications 16, 1665 (2025)
Reconfigurable Integrated High-Speed Thermal Metamaterial Pixel Arrays
Y. Zhong, X. Liu, Z. Wang, T. Huang, J. Zou, S. Lin, et al. — Nano Letters 25 (33), 12712–12718 (2025)
Reconfigurable Ultrafast Thermal Metamaterial Pixel Arrays by Dual-Gate Graphene Transistors arXiv:2506.04372
Y. Zhong, X. Liu, Z. Wang, T. Huang, J. Zou, S. Lin, et al. — preprint (2025)
Power efficient MoS2 synaptic devices based on Maxwell–Wagner interfacial charging in binary oxides first author
J. Zou, S. Lin, T. Huang, H. Liu, Y. Liu, Y. Zhong, et al. — 2D Materials 11 (1), 015009 (2024)
Physical Vapor Deposition of High-Mobility P-Type Tellurium and Its Applications for Gate-Tunable van der Waals PN Photodiodes
T. Huang, S. Lin, J. Zou, Z. Wang, Y. Zhong, et al. — ACS Applied Materials & Interfaces 17 (1), 1861–1868 (2024)
Artificial neuronal devices based on emerging materials: neuronal dynamics and applications
H. Liu, Y. Qin, H.-Y. Chen, J. Wu, J. Ma, Z. Du, N. Wang, J. Zou, et al. — Advanced Materials 35 (37), 2205047 (2023)
Circuit-level memory technologies and applications based on 2D materials
J. Ma, H. Liu, N. Yang, J. Zou, S. Lin, Y. Zhang, et al. — Advanced Materials 34 (48), 2202371 (2022)
Controlled growth of two-dimensional InAs single crystals via van der Waals epitaxy
J. Dai, T. Yang, Z. Jin, Y. Zhong, X. Hu, J. Zou, et al. — Nano Research 15 (11), 9954–9959 (2022)
Strongly surface state carrier-dependent spin–orbit torque in magnetic topological insulators
X. Che, Q. Pan, B. Vareskic, J. Zou, L. Pan, P. Zhang, et al. — Advanced Materials 32 (16), 1907661 (2020)
Band engineering in epitaxial monolayer transition metal dichalcogenides alloy MoxW1−xSe2 thin films
X. Xie, Y. Ding, J. Zong, W. Chen, J. Zou, et al. — Applied Physics Letters 116 (19) (2020)
Growth and Thermo-driven Crystalline Phase Transition of Metastable Monolayer 1T′-WSe2 Thin Film doi:10.1038/s41598-019-39238-7
W. Chen, X. Xie, J. Zong, T. Chen, D. Lin, F. Yu, S. Jin, L. Zhou, J. Zou, et al. — Scientific Reports 9, 2685 (2019)
Full list and citation metrics on Google Scholar.
The fastest way to reach me is email: jingyizou01@gmail.com, or find me on LinkedIn. I'm always happy to talk about tactile sensing, robot learning, or low-cost robot hardware — and I'm currently open to postdoc and industry R&D opportunities.