AI Engineer · Machine Learning Engineer — I build systems with Large Language Models (LLMs), Retrieval-Augmented Generation (RAG), and AI agents, plus the automation around them — and I ship them to production.
I took the long way here — electronics (CETI) → robotics & mechatronics engineering (UDG) → industrial automation → full-stack development → data & machine learning → AI and LLM products — and it converged into one strength: building complete systems end-to-end and keeping them running.
Guadalajara, Mexico (US Central time)
Languages
Frameworks & AI
Tools & Cloud
Maity — AI communication-training mentor for sales teams (co-founder & CTO)
Live on web (Next.js) and desktop (Rust/Tauri); mobile in development (Flutter). Spanish-language product; the codebase is private — it's the company's core product. I build the AI layer: LLM and agent workflows on the OpenAI and Anthropic SDKs, RAG with semantic search over pgvector, voice/wearable audio → cloud transcription, and the data infrastructure (Azure, Supabase/PostgreSQL).
volve-oil-production-databricks — machine learning pipeline, honestly evaluated
End-to-end pipeline on Databricks: PySpark, Delta Lake, MLflow (autolog), Unity Catalog. The part I care about most: a random train/test split reported R² = 0.99; a temporal split reported R² = −2.83. I ruled out feature-level leakage first (removing the suspect features changed nothing), then traced the inflated number to the evaluation setup itself — the random split let the model interpolate between known days instead of extrapolating to future ones — and published the temporal result as the honest baseline, with the full three-iteration investigation in the repo.
| Project | In one line |
|---|---|
| CUANTY · repo | Production ERP with fully automated Mexican e-invoicing (CFDI 4.0: XML sealing + PAC integration), built and deployed for one client. Next.js · TypeScript · Supabase. |
| Client automations (private) | LLM content automation for a consulting client — AI-generated video, automated LinkedIn publishing, an automated WordPress blog — built with Make (workflows running in production) and n8n. |
| Self-hosted helpdesk (private) | Zammad deployed for a client on their own infrastructure — Docker Compose + Cloudflare Tunnel (zero-trust access), running in production. |
| vehicle_ads_eda | Streamlit EDA dashboard — live demo (free tier: allow ~60 s cold start). TripleTen coursework. |
| Oil_Well_Selection | Profitability optimization under a budget constraint — bootstrap resampling (1,000 iterations) to quantify risk before recommending. TripleTen coursework. |
| Game_Success_Insights | Video-game market EDA + statistical hypothesis testing across platforms and genres. TripleTen coursework. |
Currently: TripleTen Data Science bootcamp (in progress) · building out my formal MLOps tooling layer — Docker + CI/CD around a deployed model with monitoring.
jagv.1390@hotmail.com · LinkedIn
Open to remote AI Engineer and Machine Learning Engineer roles, and to freelance LLM and automation projects.