Gustavo Aispuro
ES gus@nolu.ai

AI Solutions ArchitectFounder of NOLUCuliacán, MX · remote

Next-gen software doesn't answer questions, it does the work.

I'm Gustavo Aispuro. I build AI systems that run inside real businesses, and I run my own with a team of about thirty agents. That floor is how it works. Go ahead, poke a robot.

Take the tour
  • 9,000+ hours a year automated for clients
  • $150K+ USD in operational impact
  • 5+ enterprise clients shipped in under a year

Claude Certified Architect, Professional Certified in the Claude Partner Network. M.Sc. capstone built with MIT CTL

Portrait of Gustavo Aispuro
ID 001 · HUMAN Gus The human in the loop Nothing ships without his yes.
Chief of staff 29 on shift. I route, they draft, Gus says yes.

One human. About thirty agents.

This is how I run my own businesses. A chief of staff agent takes whatever comes in and routes it to a specialist. The specialists draft, convert, research and prepare. Nothing is sent, posted or bought without my explicit yes.

  1. 1
    A task arrives A file, a lead, a listing, a question.
    0
  2. 2
    It gets routed The chief of staff picks the squad by prefix.
    0
  3. 3
    A specialist does the work Converts, drafts, researches, prepares.
    0
  4. 4
    It waits for a human yes Every send, post or purchase stops here.
    0
  5. 5
    Then it ships With the yes on record, it goes out.
    0

Counters run since you opened this page.

Be Gus for a second

0 drafts waiting at the gate

Four squads, one prefix each

NOLU the consultancy

ETERNNA the jewelry brand

INGRESOS income projects

HERRAMIENTAS tools

Plus about twenty more under the same four prefixes.

Right now

Claude Code, Codex and Cursor running in parallel, including cloud agents from my phone. About 500M tokens a day. On the TokenMaxxer leaderboard as @gucci.

From meeting to proposal

Plaud records the meeting and the transcript goes to the agent team. Some pick out what the client asked for, some start building, some quote. Obsidian holds the context. What used to take days or weeks now gets resolved in half an hour.

The floor is a scripted simulation of a real workflow. Nothing on this page is live and no model is called.

Casebook 3 kinds of work on file. Client names redacted, by policy.

NOLU builds AI that does the work.

An AI consultancy for mid-market and enterprise companies, mostly in Mexico. The job is to take AI from a demo to something that runs every day. Custom solutions powered by Claude, from discovery to deployed product, on infrastructure the client owns. I lead five engineers across parallel deployments.

Three kinds of work

  • 01
    Family office An operations platform the client fully owns

    WhatsApp intake, Claude task classification, role-based portals and default-deny security from day one. It grew into a multi-project engagement with an on-premise local AI for sensitive data.

  • 02
    Logistics A 100+ vehicle fleet you can ask questions

    An operations dashboard with an embedded Claude assistant for natural-language queries, a delivery-evidence validation flow, and invoice reconciliation built on n8n with automated PDF extraction.

  • 03
    Enablement AI training for 4 companies, 150+ people

    Hands-on programs from front-line staff to C-level, plus advisory work.

Also delivered

  • Automated bank statement processing for an automotive group
  • Bank reconciliation automation for a finance team
  • A documentation and executive reporting automation platform
  • Performance and rest monitoring for a transport company
384+

Where it started

My first production automation was an email-to-invoice reconciliation pipeline at a logistics company. It removed 384+ hours a year of manual accounting. That project became NOLU.

Certified in the Claude Partner Network. nolu.ai

Raíz Scope first, build second. Both methods are written down.

How the work gets scoped.

NOLU runs on two internal methods, used on every engagement.

RAÍZ

Finds the automation worth building. Five phases and a readiness score.

  • 01
  • 02
  • 03
  • 04
  • 05
  • Score
CRIBA

A sieve for what a client says they want, down to what will hold up in production.

  • Capturar
  • Revisar
  • Impacto
  • Balancear
  • Asignar

Rules that do not bend

  1. Fixed steps go in code. Judgment goes in an agent. Never hand a deterministic problem to a non-deterministic system.
  2. No data, no project. If it cannot be measured before, nobody can prove it improved after.
  3. The LLM never decides amounts. Money, quantities and approvals stay deterministic. The model reads and drafts.
  4. Meet the operation where it is. Half of these systems run through WhatsApp, because that is where the work happens.
  5. The client owns the infrastructure. No dependency on NOLU to keep the lights on.

What I build with

LLM systems
Claude agents and assistants in production, retrieval over private corpora, structured extraction from invoices, CFDI XML and scanned PDFs.
Application
TypeScript and Next.js front to back, Python for data and extraction, Postgres when the project needs it.
Channels
WhatsApp Business API as a first-class interface.
Discipline
Test-first: behavior in Gherkin before code, coverage and mutation thresholds in CI, default-deny auth before anything ships.
Workshop None of this was scoped by a client.

What I build when nobody is paying.

  • Master's capstone · built with MIT CyberRadar

    Cybersecurity foresight for supply chains, built with the MIT Center for Transportation & Logistics. It places emerging cyber risks on a ten-year radar, links each one to similar past incidents, and ties every claim to an archived source and a verbatim quote.

    ctlradar.com
  • Agents Prediction-market swarm

    A multi-agent system that researches, argues and places positions.

  • Agents Six-role dev team

    Analyst, PM, Architect, GitHub Ops, Developer and QA, with model routing across Haiku, Sonnet and Opus.

  • Hardware Voice assistant

    Whisper, Claude and ElevenLabs, running on hardware I wired myself.

  • Product Cordex

    A company brain: it connects a company's tools into one searchable memory. Selected for the Hyperagent Founding 500 with $20K in credits.

    gocordex.com
  • Hackathon · 3rd place BREACH

    An autonomous security auditor. It attacks an app from the outside every day and learns from every vulnerability it misses. 3rd place at the OpenAI × Kavak × Primero RSI Hackathon in Mexico City: it went from 12 to 51 flaws found in a day, with zero false alarms.

    trybreach.com
    Gustavo working on a laptop at the hackathon, with Mexico City behind the window
    Heads down at the hackathon.

Pointed next at edge AI and TinyML.

Teaching

I publish as @gus_ia, demystifying AI for Latin American founders. Professor tone, never guru.

One more thing

Fine jewelry Eternna Joyería

A fine jewelry brand I co-founded, run separately from NOLU. Agents run parts of it: production uploads, a design Hub and client quotes.

eternna.com.mx
Archive The résumé version. Same facts, less scrolling.

The CV.

I help companies move AI from a demo to something that runs every day. I started on the business side, auditing and standardizing processes inside companies. Now I build the agents that automate them.

Full profile on LinkedIn

Experience

  1. Jan 2026 to now
    Founder & AI Solutions Architect NOLU
    • AI consultancy for mid-market companies in Mexico. We map how the work actually gets done, find where AI pays, and build the system end to end, from discovery to ongoing support.
    • Shipped: an operations platform for a family office (WhatsApp intake, Claude task classification, role-based portals, on-premise local AI), a dashboard with an embedded Claude assistant for a 100+ vehicle fleet, bank statement processing for an automotive group, bank reconciliation for a finance team, and executive reporting automation.
    • AI training for 4 companies and 150+ people, from front line to C-level.
    • 9,000+ hours a year automated and $150K+ in operational impact across 5+ enterprise clients in under a year. Leads five engineers.
  2. Mar 2025 to Jan 2026
    Continuous Improvement & Quality Auditor Happy Mile · logistics and transport
    • Audited processes against the standards of industry leaders, standardized policies and procedures, and implemented improvements with technology.
    • Built the company's first production automation: an email-to-invoice reconciliation pipeline that removed 384+ hours a year of manual accounting. It became the seed of NOLU.
  3. Oct 2024 to Mar 2025
    Operations Training Program Happy Mile
    • Rotated through every area to learn the operation and the administration from the inside.
  4. May to Aug 2024
    Monitoring Analyst, internship Happy Mile
    • Monitored safety and protocol compliance for 50+ units nationwide and handled incidents with drivers on the road.
  5. Feb to Jun 2023
    Consulting Intern Infocus Consultores
    • Process design, standardization and audits.
  6. 2021 to 2022
    Barista, then Administration Wolfshack Coffee and Tea · part time
    • Started behind the bar and ended up managing three local coffee shops: staff, resources and whatever broke that day.

Education

  • M.Sc. Applied Artificial Intelligence Tecnológico de Monterrey 2024 to 2026

    Capstone: CyberRadar, built with the MIT Center for Transportation & Logistics. ctlradar.com

  • B.B.A. Strategy & Business Transformation Tecnológico de Monterrey Aug 2020 to Jun 2024
  • Exchange semester EU Business School, Geneva Oct 2023 to Jan 2024
  • Concentration on Digital Transformation Stanford Engineering, through Tecnológico de Monterrey Completed Nov 2025
Gustavo signing his master's diploma at Tecnológico de Monterrey, pen raised high above his head
Signing my master's in AI. Yes, I am exaggerating for comedic effect.

Certifications

  • Claude Certified Architect, Professional Anthropic · Nov 2026
  • Claude Certified Architect, Foundations Anthropic · Jun 2026
  • Claude Partner Badge, Claude Code Anthropic · Jul 2026
  • Introduction to Model Context Protocol Anthropic · Mar 2026
  • Google Cloud Computing Foundations Google · Jan 2026
  • Scrum Master

Programs and talks

  • ICAMI, PSM executive program (in progress)
  • AI & Innovation workshop, Arizona State University
  • OpenAI DevDay Exchange, Mexico City
  • Panelist at Semana Tec, Tecnológico de Monterrey, Campus Sinaloa

Languages

  • Spanish, native
  • English C1, TOEFL iBT 109
  • French, elementary

Works with

  • AI agents
  • Multi-agent orchestration
  • Generative AI architecture
  • Solution architecture
  • Business process automation
  • Context engineering
  • Claude API and Agent SDK
  • Claude Code
  • Python
  • SQL
  • TypeScript
  • n8n
  • Make
  • WhatsApp Business API
  • Tableau

Work authorization

Eligible for TN (USMCA) and H-1B sponsorship. Open to U.S. relocation.

Bridge Line open.

Let's talk.

Email is the fastest route. English or Spanish, engineers or executives.

What I think I am good at

Business I can read the operation.

A degree in strategy and business transformation, a job auditing and standardizing processes inside a logistics company, and a consultancy I sell and run myself.

Technical I can build the system.

A master's in applied AI with a capstone built with MIT, certified by Anthropic as a Claude architect, with agents running in production: retrieval, document extraction, WhatsApp as the interface, test-first code.

Where they meet I turn a vague complaint into something that ships.

I can sit with a CFO to understand the problem and then deliver the agent that solves it. Projects usually die between what the business says and what gets built. That gap is where I work.

Culiacán, Sinaloa, México. Remote with US and global teams.