Veronica with friends overlooking a city at dusk
Veronica with a friend in Yosemite
Veronica with friends holding snacks
Veronica as a child in sunglasses
Veronica with a friend at home
Veronica with a friend in a grand hall
Veronica at a concert with a friend
Veronica at dinner
Veronica at graduation with family
Veronica with a friend in a lobby
Hazy blue mountains under a bright sky
A city skyline reflected in a river at night
Veronica with a friend in Yosemite
Veronica with friends holding snacks
Veronica as a child in sunglasses
Veronica with a friend at home
Veronica at a concert with a friend
Veronica at dinner
Veronica at graduation with family
Veronica with a friend in a lobby
Hazy blue mountains under a bright sky
A river and bridge under a pink dusk sky
Sun rays falling on islands in a calm sea

physical ai · world modeling · gaming

about me

Hi! I'm Veronica, a CS student at Stanford. I'm drawn to problems relating to visual perception, stemming from my interests in both computer vision and art. I'm especially interested in using diffusion and generative AI to build realistic world models to use in fields like robotic training, art, games, and HCI. I'm currently conducting robotics research at The Movement Lab at Stanford, where I'm advised by Takara Truong and Dr. Karen Liu. I've also conducted research for surgical video understanding (with Dr. Jinendra Ekanayake from Stanford School of Medicine) and quantum computing (with Dr. Thomas Cohen and Hyunwoo Oh at the University of Maryland). In the summer of 2026, I did an internship at Boeing where I worked on operating system design for the P-8 aircraft and helped deploy local LLMs. Outside of research, I love spending time on graphic design and studio art.

education

  • Stanford University

    B.S. Computer Science, AI TrackStanford, CA

    2025 – 2029

experience

  • Copilot Innovation Fellow Remote

    at Microsoft

    October 2026 – Present

    • Directly working with product and research teams to test and launch early stage AI products.
    • Microsoft Copilot
    • AI
    • LLMs
    • Product Testing
  • Student Researcher Stanford, CA

    at The Movement Lab

    advised by Takara Truong, Dr. Karen Liu

    August 2026 – Present

    • Capturing expressive mocap and retargeting it to the Fauna Sprout, cleaning up balance, foot-contact, and joint-limit artifacts.
    • Training whole-body motion trackers with PPO by fine-tuning a LoRA adapter on NVIDIA's G1 tracker, tuning rewards, terminations, and domain randomization.
    • Taking policies from Isaac Lab to MuJoCo to hardware, where Sprout tracks over a minute of a LAFAN1 dance.
    • Building agent-driven teleoperation for an AgiBot humanoid (pick-and-place, door opening) with depth and LiDAR data.
    • Isaac Lab
    • MuJoCo
    • PPO
    • LoRA
    • Motion Retargeting
    • Motion Capture
    • Sim-to-Real
    • Domain Randomization
    • ROS 2
    • LiDAR
  • VR Software Engineer Stanford, CA

    at Stanford Graduate School of Education

    September 2026 – Present

    • Developing virtual experiences with Unity and Blender MCPs for the Futuring for Equity Lab on the Quest 3.
    • Unity
    • Blender
    • MCP
    • Meta Quest 3
    • VR
  • Software Engineering Intern Seattle, WA

    at Boeing

    June 2026 – August 2026

    • Worked on operating system design for the P-8 aircraft mission computer. Validated kernel crash persistence and designed a live serial logging system, allowing states and logs across 45 systems to be preserved.
    • Wired active ports for Lantronix console servers and designed an approach to log each console session with tmux/telnet session management, systemd startup, and logrotate for organized data management.
    • Prototyped a locally hosted Bonsai LLM workflow for analyzing historical diagnostic logs while keeping critical data internal, saving engineers ~30% time during debugging.
    • C
    • C++
    • Python
    • Bash
    • Perl
    • Linux
    • Local LLMs
    • Mission Systems
    • Virtual Integration
  • Research Intern College Park, MD

    at The University of Maryland

    advised by Thomas D. Cohen (UMD), Hyunwoo Oh (UMD)

    publication

    May 2024 – August 2024

    • Modeled random 4x4 Hamiltonian paths and numerically solved the time-dependent Schrödinger equation to characterize the computational cost of maintaining adiabaticity, finding O(L log L) scaling w.r.t. path length.
    • Built a simulation pipeline in Python for 1k+ runs with 10% error gates and cross-validated results across 3 independent cost proxies and Hamiltonian draws to ensure the scaling behavior was generalizable.
    • Coauthored a peer-reviewed publication in European Physical Journal A.
    • Python
    • Mathematica
    • Quantum Simulation
    • Adiabatic Theorem
    • Numerical Methods

projects

  • mlvrsimulation

    Blackwell: Agentic Surgery Environments

    AI Agents Hackathon

    • Anthropic API
    • MCP
    • FastAPI
    • Three.js
    • WebXR
    • WebSockets
    • NVIDIA Warp
    • Newton
    • Firecrawl
    • Co-developed BLACKWELL, an AI-powered VR surgical simulation prototype that converts natural-language requests into interactive practice environments.
    • Combined an Anthropic-powered agent, anatomical asset retrieval, and MCP tools to assemble procedure-specific scenes with instructional annotations.
    • Enabled real-time interaction on Meta Quest 2 through WebXR and Three.js, backed by NVIDIA Warp and Newton for tissue simulation.
  • researchml

    Surgical Phase Recognition for Aneurysm Clipping

    Stanford University · CS231N

    w/ Emily Oberleitner, Nicole Wong, Dr. Jinendra Ekanayake

    paper poster

    • F1@10: 0.944
    • Best Val Acc: 95.7%
    • Edit Dist: 0.809
    • Python
    • PyTorch
    • ResNet50
    • MS-TCN
    • Transformer
    • Viterbi Decoding
    • CVAT
    • t-SNE
    • Confusion Matrix
    • Ablation Study
    • GCP
    • Collaborated with Stanford School of Medicine.
    • Sourced a proprietary dataset of 48 intraoperative microscope videos (40,725 labeled frames) of aneurysm clipping surgery, annotated using CVAT across 4 surgical phases: Brain Exposure, Parent Vessel Identification, Dome & Neck Identification, and Clipping. Used video-level train/val splits to prevent temporal data leakage.
    • Designed NeuroOperA, a causal transformer for phase recognition adapted from the laparoscopic OperA — and outperformed it, achieving a Viterbi segmental F1@10 of 0.944 vs. OperA's ~0.80. Also implemented MS-TCN (which OperA did not), achieving ~95% validation accuracy vs. OperA's 92%.
    • Showed that fine-tuning ResNet50 on surgical frames (vs. frozen ImageNet weights) was the single largest factor: frame accuracy jumped from 53% to 95.7% and F1@10 from 0.553 to 0.895.
    • Applied Viterbi decoding with a data-driven learned transition matrix, outperforming hand-crafted surgical priors across all thresholds.
  • researchml

    EvolveGCN-T: Self-Attention for Dynamic Graph Weight Evolution

    Stanford University · CS229

    w/ Victoria Yang, Kaci Morris

    paper

    • +8.4pt micro-F1 on Bitcoin-OTC
    • GRU → Transformer weight evolution
    • Reproduced baselines to ±1%
    • Python
    • PyTorch
    • Graph Neural Networks
    • Transformer
    • EvolveGCN
    • Weights & Biases
    • Scikit-learn
    • Docker
    • GCP
    • Proposed EvolveGCN-T, replacing EvolveGCN's GRU-based weight evolution with a Transformer encoder that self-attends over the explicit history of GCN weight matrices rather than node embeddings.
    • Outperformed the matched recurrent baseline (EvolveGCN-O) on Bitcoin-OTC edge classification: micro-F1 0.783 vs. 0.699, a +8.4 point improvement.
    • Reproduced published EvolveGCN baselines to within ±1% (Elliptic illicit-F1: 0.578 vs. paper's 0.51; SBM MAP: 0.194 vs. 0.199) before introducing the proposed variant.
    • Identified optimization instability as the primary bottleneck since self-attention showed no consistent benefit from longer history windows.
    • Evaluated across SBM (synthetic link prediction) and Bitcoin-OTC (signed trust network edge classification).
  • full-stackml

    JobShield: Detecting Fraudulent Job Postings

    Stanford University · Team 19

    w/ Yohannes Aklilu, Anna Roth, Anayochukwu Edwin Uche, Victoria Yang

    poster

    • F1: 0.913
    • 4.5× cheaper than LLM-only
    • 95% fraud caught
    • Next.js
    • TypeScript
    • Supabase
    • Gemini 2.5 Flash
    • Scikit-learn
    • TF-IDF
    • Vercel
    • PostgreSQL
    • GCP
    • Built a full-stack job posting platform with a three-layer fraud detection pipeline targeting real malware attack vectors (OtterCookie, FlexibleFerret) that have been active since 2024.
    • Hybrid LR→LLM pipeline achieved F1 of 0.913, catching 95/100 fraudulent postings while being 4.5× cheaper than using Gemini alone ($0.077/1k vs $0.35/1k).
    • Automatic feedback loop injects every moderator decision as a labeled few-shot example into subsequent LLM calls, enabling continuous improvement without retraining.
    • Logistic Regression + TF-IDF on posting text plus 8 metadata flags (missing salary, missing requirements, etc.) handles most cases in under 0.05ms.
  • engineeringmicrocontroller

    Chimes: Assistive Audio ID Device

    Chimes · Volunteering

    • 30.1% faster ID
    • 6-participant user study
    • ESP32
    • RFID
    • ESP-NOW
    • C++
    • Arduino
    • OnShape
    • CAD
    • 3D Printing
    • Raspberry Pi
    • Designed an RFID-based audio identification device for visually impaired custodial employees at Chimes, a nonprofit employing people with disabilities — helping workers safely distinguish cleaning chemicals without relying on color or label recognition.
    • Built a two-ESP32 pipeline: an RFID reader identifies tagged equipment and transmits the ID via ESP-NOW to a receiver ESP32, which matches it to an audio file on an SD card and plays it through a 3D-printed speaker enclosure.
    • Ran a blindfolded user study with 6 participants across 6 trials — participants identified 5 chemicals 30.1% faster with the device; iterated on the design based on results, adding tactile attachment points and upgrading to an external speaker module.
    • CAD-modeled a custom speaker enclosure in OnShape with honeycomb mesh acoustics, easy-repair access, and structural independence; housed the receiver ESP32, serial MP3 module, and speaker driver inside.
    • Device was designed to scale to hundreds of employees across Chimes locations.

technical skills

  • Languages

    Python, JavaScript, TypeScript, HTML/CSS, LaTeX, C++, C, Java, x86

  • Robotics & Simulation

    Isaac Lab, Reinforcement Learning, Motion Capture, LIDAR, NVIDIA Warp, Newton, Unity, Blender, WebXR

  • ML & AI

    PyTorch, Scikit-learn, ResNet50, Transformers, RNNs, Graph Neural Networks, MS-TCN, TF-IDF, Viterbi Decoding, Weights & Biases

  • Web & Full-Stack

    React.js, Next.js, Node.js, Tailwind CSS, ShadCN UI, Express.js

  • Data & Backend

    PostgreSQL, MongoDB, Supabase, Excel

  • Cloud & Tools

    AWS, Vercel, GCP, Git, Docker, CVAT

  • Quantum & Scientific Computing

    Adiabatic state preparation, Quantum simulation, NumPy, SciPy

  • Engineering

    OnShape, Fusion 360, 3D Printing, Woodworking, Soldering, Arduino, Raspberry Pi, CNC

  • Design

    Figma, Procreate, Procreate Dreams

art

featured in

© 2026 Veronica Wang. All rights reserved.

🎁 surprise