Research Depth.
Industry Impact.

2025 — Present

AI Systems Architect & Automation Consultant

Swiss Just · Remote

• Evolved a consultant management system through four architectural stages: local Excel/SQLite scripts → cloud ETL → REST API → full-stack web app, designing every architectural decision end-to-end.
• Migrated SQLite to Supabase (PostgreSQL), designing RLS policies and concurrent Python ETL scripts for automated weekly ingestion of sales and hierarchy data.
• Rebuilt the backend from Flask to FastAPI with async endpoints, JWT auth middleware, and clean service-layer architecture.
• Designed and directed the AI-assisted implementation of a Next.js + TypeScript frontend with role-based access control, auto-logout, and a private CRM with smart omni-field search.
• Designed a transparent churn-risk engine with human-readable rule logic and admin-configurable thresholds — applying XAI principles so team leaders can audit every algorithmic decision.
• Designed a real-time personalized messaging engine for 160+ consultants — admin-editable master templates stored in Supabase with dynamic variable interpolation and per-consultant overrides, replacing static batch generation with on-demand session-aware content.
• Developed a greedy optimization algorithm for minimum-cost product combinations ensuring consultants reach required sales targets.
• Conducted authorization testing on the web application to verify privilege escalation prevention — validating that non-admin users cannot access admin routes or sensitive data under any session state.
• Deployed and hardened the production VPS end-to-end: Docker Compose services behind a Traefik reverse proxy, exposed only through a Cloudflare Tunnel — zero public ports — with dedicated non-root service users and fail2ban on top.
• Directed a multi-agent AI workflow — Claude (Anthropic) as senior, Gemini as junior — reviewing outputs, making architectural decisions, and coordinating via structured handoff documents.

2026

AI Trainer

Outlier · Remote

• Evaluated multimodal content to determine whether images and audio were generated by AI systems.
• Assessed artistic sketches to determine feasibility for realistic reconstruction using generative AI models.
• Provided structured feedback contributing to improvements in frontier multimodal AI systems.

Modeling Image Tone Dichotomy with the Power Function

Axel Martinez, Gustavo Olague, Emilio Hernandez

• Applied the tone-dichotomy mathematical model to recover hidden structural information from complex visual imagery.
• Demonstrated applications including cultural heritage analysis and structural detail recovery.

Adversarial attacks dataset for low light image enhancement

Axel Martinez, Matthieu Olague, Gustavo Olague, Emilio Hernandez, Julio Cesar Lopez-Arredondo

Peer Reviewer

Engineering Applications of AI · Elsevier

Ensuring methodological clarity and scientific rigor for manuscripts in the field of Artificial Intelligence.

MSc · 2022 — 2024

Computer Science

CICESE · Baja California, México

Thesis: Use of Dichotomies for the Enhancement of Low-Light Colored Images

• Developed an intrinsically explainable analytical model capable of recovering hidden information from low-light images.
• Leveraged parameterized modeling and optimization techniques to maintain visual fidelity and interpretability.

Advised by Dr. Gustavo Olague — recognized among the World's Top 2% Scientists (Stanford · Elsevier, 2023).

BSc · 2017 — 2022

Biomedical Engineering

UDLAP · Puebla, México

Built foundations in signal processing, human physiology, and medical instrumentation — the bridge between biology and computation that shapes my research lens today.