Everyday
Economy Atlas
Interactive research
An interactive atlas of how people spend, work, and live, connecting economic data to questions worth building around.
RayXC / San Francisco
I started RayXC in January 2026 as my startup company entity—the name that has tracked every pivot since. I build useful, occasionally weird products for people and AI agents along the way.
Raymond XuFounder / Builder
About
Turning ideas into working software
Projects & experiments / 12
Visual coding / Waitlist
Vibe code visually. Video explainers that help you see what your code is doing.
02AI inference / Research
Exploring faster, more economical inference for AI agents, from GPU benchmarks to long-context workloads.
Interactive research
An interactive atlas of how people spend, work, and live, connecting economic data to questions worth building around.
Health care navigation
A personal care navigator for making sense of medical records, research, and the next steps in complex health care.
05Contract review / Prototype
Playbook-driven contract review with native Word tracked changes and plain-English comments explaining each edit.
06Agent marketplace
Product reviews by AI agents, for AI agents — Yelp for the agent internet.
07AI infrastructure
A workspace for AI data center site readiness, from power and permits to predevelopment decisions.
08AI workforce
Self-learning AI employees that learn a team's workflows and get better with use.
09Shared intelligence
A shared AI chat that learns from everybody talking to it.
10Web creation
Really, really fast vibe-coded websites.
11AI experiment
A vibe-coded AI model you can talk to called “bruh.”
12Please help
An internet-native plea for assistance with Raymond's ankle.
The build log
Three years of building, from real-estate software to AI agents, infrastructure, and visual coding. The chapters change; the building keeps going.
Loading the build log…
Previously / 2018–2026
The companies, products, and research I worked on before starting RayXC in January 2026.
Experience on LinkedIn ↗–
Founder at Modern Realty · YC S24
Built an AI home-buying agent handling texts, email, tour scheduling, and property research. Scaled to over $500k in total real-estate commission revenue. The company is now Mage Legal ↗.
–

Founder
Built an automatic summarizer for real-estate disclosure documents.
–
Machine Learning Engineer
Prototyped generative AI products with Stable Diffusion and optimized LoRA training for personalized models.
–
Machine Learning Software Engineer
Improved voice-command understanding on the NLP retrieval team through BERT embeddings and ML experiment pipelines.
Built entity-extraction models and ML infrastructure for forms and invoices, including spatial features, fine-tuning, and tokenization experiments.
–
Team Lead & Acting Program Manager · Part-time
Coordinated outreach to small businesses for petitions to local mayors.
Coordinated Humanity CALLS volunteers for voter outreach.
–
Founder
Built a web-research tool that surfaced potential sources and related content when you clicked a paragraph.
–
Founder in Residence
Developed an NLP product for text recommendations on websites during the founder residency in Brooklyn.
–
Machine Learning Engineer
Developed infrared detection, segmentation, and orientation models for small-drone navigation. Led low-light underwater imaging and depth research, and built annotation and data-augmentation tools.
Selected repositories
Helps AI coding tools find the file that is currently executing.
GH / 02A blend of the Claude Code SDK with Cloudflare agent code mode.
GH / 03Autonomous code optimization using LLMs and any computable metric.
GH / 04Personal zshrc and vimrc configuration for a fast terminal-first workflow.
GH / 05Smart test selection from AST-level code dependency analysis.
GH / 06A working collection of examples, experiments, and useful recipes.
“Simplicity is an advantage but sadly complexity sells better.”