Néstor Martínez
Néstor Martínez

AI Implementation Consultant

Available now

IbuildtheAIsystemseveryoneelsejustpitches.

I design and operate multi-agent AI systems in production for real businesses.

View my workGet in touch
numen — pipeline.sh
$ query: "what does emotional detachment mean?"
→ classify: complex / practice
→ retrieval: hybrid (exact SQL + pgvector)
→ rerank: two-stage · evidence pack: 6 passages
→ gap_check: coverage OK
→ answer
50%→90% RAG ACCURACY248 ADVISORS~300 CONV/MO90% TIME SAVED200+ ORCHESTRATED NODES9 AGENTS50%→90% RAG ACCURACY248 ADVISORS~300 CONV/MO90% TIME SAVED200+ ORCHESTRATED NODES9 AGENTS
// numen — production architecturelive pipeline
WhatsAppinbound msgRouterintentIntentRAG FormatRAG ExecAppointmentInsistenceRe-introRhythmMsg AnalysisFAQRAGHíbridoBM25 + vecRerankercross-encoderRespuestato advisor
90%RAG accuracy
248advisors on one platform
7–10hhours/week recovered

Selected work

05 systems — 2024–2026
01

Numen

Precision RAG assistant on WhatsApp over the complete works of Dr. David R. Hawkins (~2,000 pages). Hybrid retrieval, two-stage reranking, self-learning loop.

50% → 90% RAG accuracy on WhatsApp

02

Sales Advisory Platform

CRM + RAG copilot for a large European financial advisory network that ran on spreadsheets. Personal WhatsApp AI agent living in each advisor's own number.

248 active advisors, CRM + RAG copilot

03

Gaea Legal

AI front office for an immigration law practice: WhatsApp agent, email automation and an AI social pipeline. Learns from every human correction.

~300 conversations/month, 90% content time saved

04

ListingKit

Self-serve content engine for real estate: one property listing generates six ready-to-publish assets in ~10 minutes. Built as a replicable vertical engine.

1 listing → 6 marketing assets in ~10 min

05

ACA Global Academy

Custom Skool-style learning community and CRM for a trading academy relaunching from zero. Owned platform, no per-member SaaS fees.

100-200 active students, Skool-style platform + CRM

// public repositories

Open Source

n8n-workflows-mcp

Turn your deployed n8n workflows into tools your AI agents can call — without rewriting anything.

  • TypeScript
  • MCP
  • n8n
  • zod

105 tests · CI green · npm v0.2.0

View on GitHub

llm-quorum

Route AI tasks to the right model CLI, with automatic fallback and cross-model consensus review.

  • TypeScript
  • CLI
  • Claude
  • Codex
  • Gemini

161 tests · CI green · npm v0.1.1

View on GitHub

ragproof

Self-correcting RAG answers, plus a CI eval gate that blocks the merge when hallucination rate crosses your threshold.

  • TypeScript
  • CLI
  • LLM-as-judge
  • zod

124 tests · CI green · npm v0.1.0

View on GitHub

castloop

A full AI content loop — research, script, video, publish, analytics — where every agent handoff is judged for fabrication and a human gate guards the publish.

  • TypeScript
  • Remotion
  • Whisper
  • Postiz
  • LLM-as-judge

153 tests · CI green · npm v0.1.0

View on GitHub

instinctkit

Starter instincts plus the production guardrails that keep long agent sessions from failing silently — stall detection, foreground-delegation limits, anti-assumption edits.

  • TypeScript
  • Claude Code
  • hooks
  • Sinapsis

95 tests · CI green · npm v0.1.0

View on GitHub

ListingKit

One property listing in, six on-brand marketing assets out — a vertical content engine for real-estate agencies, open sourced end to end.

  • TypeScript
  • Next.js
  • Prisma
  • Playwright
  • FFmpeg

265 tests · 9 E2E · CI green

View on GitHub

I'm an AI Implementation Consultant — I embed with a business, find where AI actually pays for itself, and build the system end to end, then operate it. My deployments run daily for teams of 5 to 250 people: RAG assistants on WhatsApp, AI-operated front offices, content pipelines.

More about me
Néstor Martínez

Let's build something that actually ships.

info@podervsfuerza.comGet in touch