RL in progress
OpenRA-RL project is actively ongoing
Building and evaluating reinforcement learning agents in the OpenRA environment, with a focus on long-horizon strategy and reliable benchmarking.
Project PageSystems Builder and Researcher
Bold products, reliable systems, and human-centered algorithms.
Hui Xu at Stony Brook University. RL, systems, and AI research.
I build and ship scalable products across global teams. I have owned and operated 10+ Linux cloud servers end-to-end (OCR pipelines, invoice verification services, databases, deployment). Algorithmic improvements reduced manual review while improving recognition speed and accuracy. I am currently focused on reinforcement learning for RTS environments and robust agent evaluation.
Reinforcement Learning
Quant Systems
LLM Research
Distributed Systems
Agent Security
Full-Stack Dev
Cloud Ops
Goal
Build great products that create value for individuals and enterprises, enabling shared wins.
Independent Ownership at Scale
I have led full-stack operations for startups, maintaining more than 10 Linux cloud servers including OCR, invoice validation, and database systems. I implemented algorithmic optimizations that reduced human QA cycles while speeding up recognition with higher accuracy.
Cross-Border Collaboration
In multinational environments, I communicate across teams and stakeholders in the US, India, and China to deliver products on schedule and with clear ownership. Delivered on-prem deployment of a full-stack business analytics platform.
News / Updates
Latest milestones, RL progress, and research highlights.
Academic service
Reviewer at Agents in the Wild Workshop (ICLR 2026)
Serving as a reviewer for agent-related submissions. Workshop date: April 26, 2026 (Rio de Janeiro, Brazil).
WorkshopCompetition
2nd Place — OpenAI-Sponsored Finance Agent Competition
Our team placed 2nd in the AgentX AgentBeats finance agent competition organized by UC Berkeley RDI and sponsored by OpenAI.
Details Our StoryResearch update
MAQuA accepted to EACL 2026 (Main Conference)
Adaptive questioning for multi-condition mental health assessment.
arXivWhat I Am Building Now
Applied systems, research, and reliable AI infrastructure.
OpenRA-RL: Reinforcement Learning for RTS
I am developing RL workflows for real-time strategy game environments. The current emphasis is on training stability, map generalization, and reproducible evaluation.
- Win Rate - track progress against scripted and learned baselines.
- Sample Efficiency - improve learning speed under limited interaction budgets.
- Generalization - verify robustness across maps, opponents, and game settings.
Status: in progress (2026).
Project PageApplied C++ in Quantitative Trading
Building scalable algorithms, backtesting tools, and market simulators.
I implement data-driven statistical arbitrage strategies and backtesting systems in C++. If I have the chance, I will explore and use the CUTE DSL to optimize GPU kernels.
- Latency - minimize so the system adapts to market conditions.
- Accuracy - keep MSE low so model quality stays strong.
- Throughput - handle as many requests per second as possible.
LLMs for Counseling, Recommendations, and Advertising
MAQuA enables assessment of multiple mental health conditions with a minimal number of adaptive questions. Current language models do not model the human subconscious. I am researching how to represent it, including L'inconscient inculqué à mon ordinateur (Book, French Edition) and Lacan and subjective topology.
Accepted to EACL 2026 (Main Conference Poster). An additional NLP research project is ongoing.
PaperAgent Security
LLM-generated code often misses complex system constraints and deep optimization. When junior engineers rely on it without strong systems or security knowledge, they can ship small-scale, non-scalable apps with weak threat modeling and unpatched vulnerabilities. I focus on making LLM agents safe, scalable, and resilient.
RepoLLMs from Scratch
What I cannot create, I do not understand.
RepoLeetCode Algorithms
700+ problems solved. Preparing video walkthroughs for challenging problems to build skill in reading, verifying, and improving LLM-written code for correctness, efficiency, and safety.
Distributed Systems Reading Lab
Paxos + 2PC, Raft + 2PC, PBFT + 2PC. These systems are still beyond what any LLM can reliably implement end-to-end from a prompt. I build them with DistAlgo (compiled to Python) for transaction workflows, and I plan to explore coroutine-based designs to reach ~1000 transactions per second per cluster. I also plan to read the papers and record tutorials.
RepoPublications
Research outputs and works in progress.
MAQuA: Adaptive Mental Health Assessment
Accepted to EACL 2026 (Main Conference). Preprint on adaptive questioning for multi-condition mental health assessment.
arXivAcademic Service
Peer review and community contribution.
Reviewer - Agents in the Wild Workshop (ICLR 2026)
Reviewing submissions on trustworthy, safe, and effective agent systems for the workshop held on April 26, 2026 in Rio de Janeiro.
Workshop SiteArt
Paintings and sketches by me.
I am a lesbian; I value the Statue of Liberty's ideals of freedom, democracy, hope, and inclusion, alongside bold exploration backed by safety.
Writing
Personal essays on Lacanian psychoanalysis (Chinese).
Selected Work
Earlier projects that shaped my product and engineering craft.