Stanford CS / AI systems / LLM agents / infrastructure
I am a Master's student in Computer Science at Stanford University, focusing on AI and systems, with recent work across LLM agents, backend infrastructure, and scalable framework design.
About
I am a Master's student in Computer Science at Stanford University, focusing on AI and systems. My recent industry experience includes a backend engineering internship at Stripe, and I will be returning to Stripe's Risk Capabilities group as a full-time engineer. On the research side, I co-develop ReCAP, a recursive reasoning and planning framework for large language models.
During my undergraduate studies in Game Design and Computer Science at NYU, I founded Kromo Games, a student-led studio where I led the development of Mikrocosmos, a multiplayer game with over 25,000 downloads on Steam. To support its growth, I built MikroFramework, a modular game framework inspired by MVC principles that emphasized scalability and extensibility. This early experience in systems and framework design continues to influence how I approach engineering challenges today.
I am motivated by problems at the intersection of systems and intelligent systems, and I enjoy building infrastructure and platforms that enable breakthroughs in AI and data-driven applications.
Selected work
Co-developed a hierarchical framework for long-horizon LLM agents that plans ahead, executes the first subtask, recursively decomposes composite work, and re-injects the parent plan on return to keep multi-level context coherent.
Optimized a core database interaction by adding a pre-check that reduced daily write transactions by over 85%, then drove a synchronous execution refactor that increased throughput under load.
Write-path optimization / execution architecture / production-safe rollout
Led a 6-person team building an online multiplayer strategy-action game with Unity, Mirror networking, PlayFab/Azure backend services, and hot-update driven content systems.
Research
My research work sits near the boundary between model capability and the systems that make those capabilities usable in longer tasks.
Designed core inference architecture for recursive, context-aware agent execution: plan-ahead decomposition, structured parent-plan re-injection during backtracking, bounded context management, and benchmark evaluation across embodied, fact-checking, and coding tasks.
Built a Transformer-enhanced relational graph convolutional model to capture temporal and spatial dependencies in unmonitored water networks, improving over GCN and RGCN baselines.
Developed a Unity-based haptic simulator with surgical-step replication, educational tools, and 3D Systems Touch device integration; awarded NYU Dean's Undergraduate Research Fund.
Experience
Course Assistant for CS224W and CS246 / Research Assistant
Core contributor to large-scale graph ML and data mining courses under Prof. Jure Leskovec, including assignments, exams, office hours, and course design, while researching LLM-based reasoning systems with Jiaxin Pei and Alex Pentland.
Returning Full-time Engineer, Risk Capabilities / Engineer Intern, Rules Infrastructure
Returning to Stripe's Risk Capabilities group full time after reducing database write load, simplifying execution architecture, and validating production changes with feature flags and monitoring during the internship.
Backend Development Intern
Built logistics prediction systems and improved a complex table query/pagination flow by 60% using TypeScript, NestJS, SQL Server, and TypeORM.
Co-founder, Lead Game Developer
Led a small team shipping Mikrocosmos, integrating multiplayer networking, PlayFab/Azure services, rapid feedback cycles, and extensible game systems.
Project archive
NeurIPS 2025 framework for recursive planning, bounded context management, and parent-plan re-injection in long-horizon LLM agents.
Transformer-enhanced RGCN for spatial-temporal prediction in unmonitored networks.
Read writeupMVC-oriented Unity framework with event systems, code generation, IOC, pools, hot-update, and resource management.
GitHubMultiplayer trading-and-combat party game with networking and backend service integration.
SteamEarlier indie game project published on itch.io, part of the undergraduate game-development period that led into systems tooling work.
itch.ioLegacy collection from my earlier personal site, covering game-design projects and prototypes from my NYU period.
Old game projects pageToolkit
Python, C++, C#, TypeScript, JavaScript, Java, SQL
PyTorch, PyG, Triton, Hugging Face Transformers, scikit-learn, Pandas, NumPy, NetworkX, OpenCV
Node.js, NestJS, SQL Server, Spark, Kafka, Docker, AWS, GCP, Azure, Grafana, Splunk, CI/CD
Unity, Mirror, PlayFab, MVC architecture, hot-update systems, shaders, haptics
Contact