About
Continually exploring new ways of expressing ideas through design and code.

The bigger picture
Turning complex ideas into working systems.
I’m a builder, an artist, and a thinkerer working at the intersection of software, operations, and intelligence. I’ve led engineering teams, shipped product across logistics, sales, healthcare, and education, and spent the last few years building tools that help small teams act with the clarity of much larger ones.
Before turning 30, I scaled a COVID-19 testing platform from 1,000 to 100,000+ tests per month for a Spanish laboratory across Mexico and the Dominican Republic, ran a 60-person distributed sales team for a US client from two operators in Monterrey, and led growth experiments that returned 1,876% on ad spend. The pattern under all of it: the same operational shape, in different industries.
Now I split my time between building product, publishing research, writing, and making cinematic short films, purposely trying to answer: how does intelligence become usable, and how do you prove it? The research half is public: pre-registered experiments on production agent systems, with the first paper published as an open preprint.
Today I work as a senior AI engineer. I build artificial intelligence systems that run in production, the evaluation that decides what ships, and the tools and data underneath them. I live in San Pedro Garza García, in the Monterrey metropolitan area, and work with teams in Mexico and the United States.
Open to senior AI engineering roles and contract engagements, remote in the US and in Mexico. CV · Engineering guides and evidence · Contact
The personal side, from how to work with me to what the personality tests say, is on Personal.
In other people’s words
Rashid is one of those people you can have a genuinely technical conversation with. He gets into the details, asks good questions, and usually sees things from an angle you hadn’t considered.
Daniel ReyesRashid is the kind of engineer I’d want involved early in a complex project. He’s thoughtful about architecture, knows where things tend to break, and isn’t afraid to question an approach that doesn’t make sense.
Michael TorresReally sharp engineer. I’ve always appreciated how Rashid thinks through problems, especially when it comes to AI systems and architecture. Definitely someone worth having a conversation with.
Andrea MitchellThreads
Four ways into the work.
Architect Building systems that make complex data feel like conversation.
I design the layer that turns disparate operations into coherent ones, then ship it. Platforms (AffiHub, AyudaLocal, Crawlio, Mentu, Vaiven, Cotizera, Abastia, Fullgen) and open-source infrastructure (Notion to Site, Airtable MCP, MetaMCP, Mentu Protocol). The pattern under all of it: cognitive infrastructure.
Go to Code
Thinker Naming the operating layer of attention.
A few hundred essays, three books on cognitive infrastructure, and a working paper with my father on lean theory applied to information systems. Writing is where the frameworks get drafted before they become products.
Go to Writing
Artist Civilizations the present has forgotten.
Recreating Stories. Cinematic short films made with historical research, direction, and a curated AI stack. Tang Dynasty, Persepolis, sacred cities, ancient battles. Memory treated as a craft.
Go to Films
Operator The work that produced the frameworks.
A COVID lab scaled past 100,000 tests per month. A 60-person remote sales team run by two operators in Monterrey. A 1,876% ROI affiliate funnel. Each case study written from inside, not after.
Go to Studies
Now
What I’m working on this season.
- Agent researchPre-registered studies of how AI agents remember and retrieve, run against production systems. First paper published: Reading More, Finding Less, with data and harness shipped for re-running.
- Engineering economicsStrategic Technical Debt ↗ prices deliberate technical debt as a call option on the validated product. Preprint on arXiv; the two primary tests are pre-registered with their analyzers frozen, and running them is the next act.
- MentuAn operating system for agentic work. Local-first recipes, a commitment ledger, and an MCP intelligence layer. Open source. mentu.ai ↗
- Recreating StoriesNew cinematic episodes about lost civilizations. Most recent: Tang Dynasty’s Last Beautiful Year ↗.
- AyudaLocalSolving home services in Mexico with AI. The right pro shows up, payment handled, every step coordinated. ayudalocal.com ↗
- The Architecture of Usable IntelligenceThe third book in the cognitive-infrastructure series. The recursive engine underneath the other two. Read draft →
Stance
What I work from.
Build tools that work and feel great.
Half of "feels great" is taste. The other half is keeping the seams visible so the user can repair the thing themselves.
Treat Notion status fields as unreliable.
The metadata of any system is always more flattering than the thing itself. Read the body, not the tag.
Ship the cinematic version.
A portfolio is permission to imagine. If the work is finished and the page is a list, the work disappears.
Write everything down.
Memory is a single point of failure. So is reputation. Documents survive both.
Short bio
For events and press.
Rashid Azarang is a Senior AI Engineer based in Monterrey, Mexico, and an expert in production AI agents and LLM evaluation. He builds artificial intelligence systems that run on real data, the evaluation gates that decide what ships, and open-source tools that other developers use. He publishes pre-registered research on how AI agents work, and writes in English and Spanish.
Rashid Azarang es ingeniero de IA senior en Monterrey, México, y experto en agentes de IA en producción y en evaluación de LLM. Construye sistemas de inteligencia artificial que operan con datos reales, las compuertas de evaluación que deciden qué se publica y herramientas de código abierto que usan otros desarrolladores. Publica investigación prerregistrada sobre cómo trabajan los agentes de IA, y escribe en español y en inglés.
Colophon
How this site is made.
Built with Next.js on Vercel. Essays, books, case studies, and the research paper are authored in Notion and synced to local markdown via a small sync tool I wrote (notion-to-site ↗). Titles are set in Geist Pixel and Cormorant Garamond, text in Geist and Geist Mono. The whole corpus is machine-readable at /llms.txt and /api/context.
Get in touch.
For work, collaborations, or anything you read here that you want to argue with.