Zhihong Jiang

Education

• New York University - M.S. in Computer Science (Expected Dec 2026, Merit Scholarship Recipient)
• New York University - B.S. in Computer Science, Minor in Applied Mathematics, Minor in Game Design (Sep 2021 - May 2024)
• Elite Preparatory Academy - Honor Class (Sep 2018 - May 2020)

Technical Focus

• ML/AI: PyTorch, TensorFlow, Hugging Face Transformers, SpaCy, Torchaudio, XGBoost, JEPA, CNNs, NLP pipelines
• Data/Infra: Pandas, NumPy, Dask, PySpark, ETL, large-scale processing, Git, Docker, Linux
• Web/Backend: React, Vite, JavaScript, HTML/CSS, Flask, Express, Node.js, REST APIs, JWT
• Game/Graphics: Unity, XLua, Godot, GameMaker, OpenGL

Gameplay / AI Systems Developer

AbyssalPact (Unity, C#, XLua) | Apr 2025 - Present

• Own enemy AI systems architecture for a turn-based card game with a hybrid C# + XLua runtime.
• Built enemy-turn execution pipeline (scoring, target selection, summon placement, fallback handling) for safer runtime behavior.
• Designed extensible AI tag + metadata model to scale decision logic as mechanics and card count grow.
• Added runtime-configurable AI parameters, compatibility shims, and debugging safeguards for faster balancing and staged migration.

Self-Supervised World Model (JEPA) for Sequential Prediction

NYU Deep Learning Course Research with Prof. Yann LeCun | Sep 2024 - Dec 2024

• Owned end-to-end implementation and training workflow for a JEPA-based world model for future latent-state prediction.
• Designed a recurrent JEPA variant combining image sequences with action inputs to model environment dynamics.
• Mitigated representation collapse using regularization strategies including VicReg and BYOL-style constraints.
• Scaled experiments to 2.5M exploration frames and built probe-based evaluation for generalization testing.

Engineering Assistant, AI-Powered Interview Simulation Tool

with Thomas Modern, Data Science Lead at Meta | Oct 2023 - Mar 2024

• Led core development of an interview simulation tool combining NLP and audio analysis for structured feedback.
• Built model-driven analysis components with Python, PyTorch, Transformers, Torchaudio, and SpaCy.
• Expanded evaluation pipeline with video-based nonverbal analysis and audio-based structure/terminology assessment.
• Improved usability with user-facing features and documentation.

Software Engineering Intern

Anhui Yuntai Transportation Development Limited | Jun 2023 - Aug 2023

• Contributed to a mobile operations system for bus dispatch and car-rental workflow digitization.
• Built a C++ data module for routes, revenue, trip duration, and passenger flow reporting.
• Supported route and operations analysis that contributed to an estimated 15% efficiency improvement.

CNN Application in Cancer Diagnosis

MIT Deep Learning Research with Prof. Mark Vogelsberger | Jun 2021 - Aug 2021

• Contributed to an AI-assisted cancer diagnosis application integrating model inference, UI workflow, and NLP support.
• Built and trained a CNN-based classification component reaching 89% accuracy.
• Supported usability improvements, onboarding materials, and research publication efforts.