About me.
I build AI agents that run real businesses.
What I do: I design and build multi-agent systems that replace manual business processes. Not chatbots — autonomous agents that qualify leads, negotiate payments, reconcile bank statements, and make decisions without human intervention.My focus: Latin America has a massive gap in business automation. While US companies have Salesforce and HubSpot, LATAM businesses run on WhatsApp and spreadsheets. I build the AI infrastructure that bridges that gap — agents that speak Spanish, understand local business context, and integrate with the tools people actually use.
What I've built:
- Sr. Ventas — Multi-agent sales platform. Led a team of ~10 people (developers, sales, marketing) alongside CTO and CEO. Built the sales and marketing pipeline, managed ad campaigns, and led the dev team. Technical: 90,000+ messages, 6,000+ conversations, 30% conversion rate (AI selling itself). Architecture: 14+ microservices, multi-tenant, Stripe, Google Calendar, PWA with offline + push notifications.
- Collection agents — Multi-agent architecture for automated receivables. Specialized agents for qualification, negotiation, payment processing. FastAPI orchestration, OpenAI reasoning, custom RAG.
- Bank reconciliation — Fuzzy matching + LLM reasoning to match statements against invoices automatically.
Research background: B.Sc. Computer Science, Tecnológico de Monterrey (96/100).
- Global Americans + U.S. Department of State — Research on disinformation in Latin America. Built BERT-based fake news classifiers, sentiment analysis pipelines, social media monitoring tools. International collaboration (Argentina, Venezuela, Colombia, Mexico). Published reports in 3 languages.
- Urban Information System — Satellite imagery analysis with Landsat for urban growth classification in Monterrey metropolitan area. ML models + web application for visualization.
- Mexicovid19 — COVID tracking platform, 18K+ monthly visits. Built simulations, ML algorithms for symptom analysis, real-time data visualization. Featured in national media.
- DFuture (2020-present) — Digital agency, 3 co-founders, team of developers and designers. 20+ enterprise clients: Mercedes-Benz, Aeroméxico, Nutella, Mobil, CONACYT, Tec de Monterrey, UANL, Government of Nuevo León.
- Geniebot / Techsoft Consulting (2017-2019) — WhatsApp sales agent built with Dialogflow. Conversational AI before the LLM hype. Sold to clients in Mexico for customer service, lead qualification, appointment booking.
- Crypto Artifex (2021) — NFT marketplace for art. 3D scanning of physical artworks to blockchain. 15% commission model.
- ManGo (2017) — Services marketplace app. Presented at INCmty. Rating system, in-app payments, real-time chat.
Stack: Python, TypeScript, Next.js, FastAPI, Docker, AWS/GCP. OpenAI, Llama, Claude. LangGraph state machines for complex agent orchestration.
Skills.
ML & LLMs
Full Stack
AI Agents
Infrastructure
Languages.
Projects.
Contact me.
Social Networks
Github
Spotify
Email me
Machine Learning & LLMs
Research projects (funded by U.S. Department of State):
- Disinformation detection — BERT-based classifier for Spanish fake news (89% accuracy). Social media monitoring across 6 Latin American countries. Published in 3 languages.
- Satellite imagery analysis — Landsat data for urban growth classification in Monterrey. ML models for zone detection and expansion tracking.
- COVID symptom analysis — ML algorithms for symptom data at Mexicovid19. Epidemic simulations (SIR models).
Production AI systems:
- Custom RAG pipelines — LangGraph state machines with hybrid search (BM25 + embeddings).
- Prompt engineering — Mexican Spanish sales conversations, objection handling.
- Sentiment analysis — VADER + custom Spanish lexicons for social media.
Tools: Python, TensorFlow, PyTorch, Scikit-learn, Pandas, OpenAI, Llama, Claude, Hugging Face
Full Stack Development
DFuture — 20+ enterprise clients:
- Mercedes-Benz x Mario García Torres — Digital experience for Zona Maco 2025
- Aeroméxico Copaski — Event app with real-time updates, push notifications, photo sharing
- Government of Nuevo León — Labor market analysis platform with economic projections
- CONACYT/Tec — Observatory of Sustainable Manufacturing, geospatial data visualization
- Nutella, Mobil — Interactive campaigns, raffle systems, WhatsApp integrations
- MAPAMH — Geospatial tool for humanitarian aid to migrants (health/safety services)
Other production systems:
- Mexicovid19 — 18K+ monthly visits, COVID simulations, real-time data viz
- Urban Information System — Satellite imagery analysis, ML models
Stack: TypeScript, React, Next.js, Node.js, FastAPI, Flask, PostgreSQL, MongoDB, Redis
AI Agents & Multi-Agent Systems
Sr. Ventas — Built and led the entire operation:
- Leadership: Team of ~10 people (dev, sales, marketing). Reported to CTO/CEO. Built sales pipeline and managed marketing campaigns.
- Scale: 90,000+ messages, 6,000+ conversations, 30% conversion rate (AI selling itself)
- Architecture: 14+ microservices, multi-tenant system, Docker + AWS
- Integrations: Stripe payments, Google Calendar, WhatsApp Business API
- Frontend: PWA with offline support, push notifications
- ML/AI: OpenAI, custom prompts for Mexican Spanish sales, RAG for product knowledge
Other agent systems:
- Collection agents — Specialized agents for qualification, negotiation, payment processing
- Bank reconciliation — Fuzzy matching + LLM reasoning for edge cases
Technical decisions: LangGraph state machines with LLM nodes for intelligent orchestration. Hybrid RAG (BM25 + embeddings). Sliding window memory + summarization.