agm.

SPECIMEN №01 · AI BUSINESS ARCHITECT

Architecting AI for software that has to work.

FIGHTER PILOT · STRATEGY → EXECUTION · ONE PERSON · AI-NATIVE

Cockpit portrait — Alejandro Gutiérrez Mourente, oxygen mask on, level flight
PORTRAIT №00 · COCKPIT · OXYGEN ON
↓ SCROLL

ABOUT

Companies in regulated industries lose months translating business problems into software. I compress that to weeks because I speak both languages and build end-to-end with AI.

I operate as a founder: I pick the problem, scope the product, ship it, and put it in front of users. I pair with commercial counterparts when the deal and the problem are worth it.

A decade as a fighter pilot and Operational Safety Officer taught me that systems either work under pressure or they fail people. That standard followed me into software: understand the operation first, then build what it actually needs.

WHAT I'VE SHIPPED AS A SOLO BUILDER

01.

A production document intelligence platform — four extraction engines, contract generation — that retired a €5M/year manual process and 100+ FTEs of equivalent workload.

02.

Backoffice automation for a multi-tenant enterprise platform: document extraction, ERP integrations, vehicle-operations tooling — wired into the existing legal, billing, and CRM stack.

03.

ERP integrations, field-provenance systems, and real-time conflict resolution — designed, built, and pitched to leadership by one person.

04.

Prototype-to-production cadence: working demos in days, integrated systems in weeks.

I've done this across automotive, aviation, fintech, legal, and defense. The pattern transfers because the method does: learn the domain fast enough to challenge the requirements, then use AI to build beyond them.

Full-stack development. LLM orchestration. Domain-driven design. One person, AI-native.

SELECTED WORK

Exhibits

Three in production. Two pre-production. One in alpha. One in build.

EXHIBIT A · DOCUMENT INTELLIGENCE

DocFields.ai

Vendor-agnostic document AI · live in production

"Twenty-four enterprise processors, confidence-tiered fallback, one normalized output. Most contracts stop at the cheapest tier."
STACK· Python · Document AI · Anthropic · OpenAI
STATUS· Production · docfields.ai
SCOPE· 24 processors · €5M/year manual process retired
NOYESNOYESDocument receivedREGEXRegex parser · 4 detectors~0 cost · < 50 msCritical fieldsextracted?DOC AIDocument AI · 24 processorsDNI · vehicle reg · invoices · contractsConfidence≥ threshold?VISIONVision · Anthropic | OpenAIclaude-sonnet-4 · gpt-4o · selectableUNIFYNormalize + provenancesingle shape · provider trace · signature checkStructured payload out~70% of docsstop here.LEGENDStart / EndStepFocal decisionDecisionMergeConvergent output
View specimen ↗
EXHIBIT B · CLIENT ONBOARDING

Auto Financing Onboarding

Client portal for an auto retail group · powered by DocFields.ai

"Cross-document conflict detection, append-only attestation per field. No human keystroke between document upload and the underwriting payload."
STACK· Next.js 16 · Hono BFF · GraphQL · DocFields.ai
STATUS· Pre-production
SCOPE· Auto-financing onboarding for a multi-tenant dealer group
OUR PLATFORMHTTPStRPCCALLEXTRACTEXTRACTFIELDSSIGNUNDERWRITEUSERCustomerbrowser · mobile-firstWEBClient PortalNext.js 16 · App RouterBFFHono BFFedge runtime · GraphQL outROUTERDocFields.ai Enginemulti-provider · fallbackCLOUDDocument AI24 EU processorsCLOUDVision ModelsAnthropic · OpenAIGUARDConflict Checkercross-document fieldsSTOREAttestation Logappend-only · per fieldOUTUnderwriting Payloaddecision-ready JSONNo human keystroke between document upload and the underwriting payload.Every field carries provenance: which document, which model, which confidence tier.LEGENDFocal · orchestrator / outputInternal serviceStoreExternal AI providerExternal API callUnderwriting handoffTrust boundary
View specimen ↗
EXHIBIT C · ONBOARDING ENGINE

Onboarding Questionnaire

Decision-tree document-set engine · live for over a year

"Thirty-five nodes. Thirty-four document types. Three languages. The graph decides which papers a customer must upload — and which stack to demand when the easy answer fails."
STACK· TypeScript · graph engine · i18n · DocFields.ai handoff
STATUS· Production · 1+ year
SCOPE· 35 nodes · 34 document types · ES / EN / PT
35 NODES · 34 DOC TYPES · 3 LANGUAGESperson.typeROOTNATURALLEGALnatural personAGE · NATIONALITY · DOCSlegal entityNIF · REPRESENTATIVE≥18<18spanish?FOCAL BRANCHminorTUTOR · LIBRO · AUTHYES · DNINODNI · 2 docsFAST PATHNIE · pasaporte · drivingFOREIGN · 4–6 DOCStutor + minor5–7 DOCSREPRESENTATIVENIF · representative · address4–6 DOCSOUTPUTrequired[]optional[]i18n_keyDocFields.aiLEGENDDECISIONINFO / FORMTERMINAL · DOC SETFOCAL · FAST PATH
View specimen ↗
EXHIBIT D · BACKOFFICE AUTOMATION

Business Management System

Autonomous purchase-document validation for an enterprise BMS

"Every field on the contract cross-checked against every other document in the bundle. Most files clear without an operator touch."
STACK· Multi-tenant BMS · DocFields.ai · cross-validation engine
STATUS· Pre-production · UAT
SCOPE· 100+ FTEs of projected equivalent workload
BMS · OPS PLATFORMUPLOADUPLOADPARSE BUNDLEEXTRACTEXTRACTFIELDS + SRCFEEDSAUTO-APPROVEFLAGBUNDLEClient DocsDNI · payslip · IBANBUNDLECar Docspermiso · ITV · ficha · contratoOPS UIBMS Consolebackoffice operator portalROUTERDocFields.ai Enginemulti-provider · fallbackCLOUDDocument AIDNI · permiso · ITV · contratoCLOUDVision ModelsAnthropic · OpenAI · fallbackLEDGERProvenance + Attestationsfield · doc · model · confidenceCHECKCross-Validation EngineDNI · plate · VIN · owner matchOUTCompras Decisionapproved file · or operator queueSample checksDNI on contract ≡ DNI extractedplate ≡ permiso ≡ ITV ≡ fichaVIN on contract ≡ VIN on fichaowner name on permiso ≡ sellerall signatures + dates validMost purchase files clear without an operator touch.When something disagrees, BMS gets one alert with the exact field, document, and conflict — not a stack to re-read.LEGENDFocal · check & outputInternal serviceLedger / storeExternal AIUpload / APIFlag / returnAuto-approvalTrust boundary
View specimen ↗
EXHIBIT E · VOICE COMMERCE

Askalog

Sub-second voice catalog agent · askalog.com (alpha)

"Ask in Spanish, get the car. Sub-second, deterministic, zero hallucinated prices — the unit economics of voice commerce, finally."
STACK· Pipecat · LiveKit · Deepgram · Cartesia · Postgres
STATUS· Alpha · askalog.com
SCOPE· Auto retail group · 24,000 SKUs · 24/7 sales coverage
STREAM AUDIO ~150MSPARTIAL TRANSCRIPTMCP · SEARCH(...)VALIDATE LIVE INVENTORYREAL VEHICLE IDSRESPONSE TOKENSHIGHLIGHT MATCHESVOICE + UI · ~90MS FIRST BYTEUSERCustomerbrowser · WebRTCSTTDeepgram Nova-3streaming · es-ESAGENTLLM + MCP Tools25+ inventory toolsTTSCartesia Sonic-2streaming · es-ESCTLGUI + Catalog24K vehicles · liveLEGENDFocal agentStreaming / APIValidated returnCustomer-facing outputLive ground truth
View specimen ↗
EXHIBIT F · MULTILINGUAL TRANSLATION

SmithVox

Real-time agentic translation for regulated meetings

"KUDO sells you interpreters. Wordly sells you captions. SmithVox sells you a tamper-evident multilingual transcript — in the speaker's own voice — and a chair agent that keeps the meeting from drifting."
STACK· Pipecat · LiveKit · Deepgram Nova-3 · GPT-4.1 · ElevenLabs
STATUS· In build · MVP week 6
SCOPE· 32 languages · per-speaker voice clone · hash-chained ledger
SMITHVOX BRAIN STACKRAW MIC INPER-LANG MIX OUTAUDIO + SPK IDSYNTHESIZEDTEXT + LANGTGT + VOICE IDLOOKUP TIMBRESTATE SYNCHASH-LOG TURNUSERSpeakers · NEN · ES · MN · DE …HUBLiveKit SFUper-participant tracksUSERListeners · Neach in chosen langSTTDiarization + STTDeepgram Nova-3BRAINTranslation HubGPT-4.1 + glossary RAGTTSTTS · MultilingualElevenLabs MML v2 · cloneSTOREVoice Clone Vaultinstant clone · session-scopedAGENTSession Chairdecisions · names · jargonLEDGERCompliance Ledgerhash chain · OpenTimestampsOne ingress.N egress mixes.each listener hears the room in their language · in the speaker's own voicePer-turn latencySTT ~250 mstranslate ~380 msTTS ~280 msjitter ~180 msp50 ≈ 1.09 sp95 ≈ 1.28 sLEGENDFocal · brain & ledgerInternal serviceStore / vaultExternal providerOptional / parallel agentAudio · WebRTCState syncNotarized handoff
View specimen ↗
EXHIBIT G · ABSA PIPELINE

Prism Engine

Aspect-based sentiment with sector RAG · whyrating.com (pre-production)

"Thirty-six primitives, intensity-scaled, anchor-grounded to the verbatim review text. Reputation noise becomes a queryable underwriting signal."
STACK· Python · Pydantic · Anthropic · pgvector
STATUS· Engine in production
SCOPE· 85+ classification runs · 36 primitives
FETCH UNCLASSIFIEDLOAD SECTOR VOCAB36 PRIMITIVES + ADDONSBATCH × 5 PARALLELABSA EXTRACTSTRUCTURED JSONVALIDATE + ANCHOR GROUNDPYDANTIC + UPSERTFEEDReview Streamv3 · LEFT JOINORCHClassification Pipelineasync · semaphore × 5RAGSector Vocabularyprimitives + addonsAIAnthropic LLMclaude-sonnet-4GUARDValidator + DBPydantic · v3 upsertLEGENDFocal orchestratorExternal / API callReturnCritical guard stepRAG / store
View specimen ↗

INTERMISSION

If a regulated AI build sounds like your problem, let's talk.

METHOD

Learn the domain fast enough to challenge the requirements, then use AI to build beyond them.

AUTOMOTIVE·
AVIATION·
FINTECH·
LEGAL·
DEFENSE
F/A-18 Hornet in vertical climb

PHOTO №01 · F/A-18 HORNET

BACKGROUND

Decade as a fighter pilot.

A decade as a fighter pilot and Operational Safety Officer in the Spanish Air Force. Operational discipline transferred to software architecture: systems either work under pressure or they fail people. That standard runs through every project here — the audit trails, the deterministic guards, the field-level provenance.

The same loop in both jobs: understand the operation first, then build what it actually needs. No improvisation on critical fields. No undeclared exceptions in production.

Self-portrait in the cockpit, oxygen mask on, level flight
PORTRAIT №01 · COCKPIT · LEVEL FLIGHT
Aviation Safety Management System dashboard — risk posture, SPIs, compliance overview
SPECIMEN №07 · SAFETY MANAGEMENT SYSTEM · AVIATIONRisk posture, SPIs, audits, investigations — the same operation I used to enforce, now as software.

CONTACT

Portrait of Alejandro Gutiérrez Mourente
PORTRAIT №02 · 2026

Let's talk.

I'm building AI-native products in regulated industries. If you're a founder, operator, or investor with a problem worth attacking and the network to put it in front of customers — let's talk.

Address revealed on click · spoofing-resistant