Alejandro De La Cruz

Professional Experience

NASA Space Apps Challenge — Global Nominee

Designed and implemented novel training environment for ML models to classify exoplanets using NASA astronomical datasets (3,000+ confirmed exoplanets)
Built rigorous evaluation framework measuring genuine model capability in ambiguous scenarios, going beyond standard accuracy metrics
Architected procedural world generation engine where physical variables deterministically define environment geometry, inspired by world model architectures
Owned end-to-end infrastructure: containerized ML pipeline with Docker, achieved sub-200ms inference latency despite complex 3D rendering
Selected as NASA Space Apps Global Nominee from 60,000+ participants

Remote

Manage production infrastructure for international banking clients (Banco Pichincha Ecuador, BCP), reducing incident response time by 40%
Operate distributed systems across OpenShift 4 and Azure Kubernetes Service, debugging complex microservice interactions using Dynatrace, Grafana, Kibana
Built automated incident detection pipelines querying Oracle PL/SQL and SQL Server, reducing MTTR from hours to minutes
Developed RPA automations for project management across Global Delivery Network Spain using Power Platform
Architected RPA system for critical incident notification for Alicorp (Peru's largest consumer goods supply chain)
Lead war room coordination and technical training for global teams

Remote

Delivered 6 end-to-end AI/ML and software solutions across e-commerce, healthcare, and finance
Implemented LLM-powered automation systems and generative AI applications with production reliability
Built scalable ML infrastructure using Docker Compose, React/Vue.js, Django, PostgreSQL, Nginx on Ubuntu Server

Hybrid

Developed legal document management system using Laravel and Vue.js, improving administrative efficiency by 60%
Architected RESTful API services balancing engineering rigor with rapid feature delivery

Remote

Volunteer

Owned end-to-end development of thesis project management system over 17-month period
Designed complete system architecture with comprehensive documentation and user manuals

Implementing custom Gymnasium-compatible environments for long-horizon tasks, studying reward shaping and PPO/SAC algorithms.

Learning LoRA, QLoRA techniques and implementing custom evaluation frameworks. Studying RLHF and alignment research.

Deep-diving into JEPA architectures and connections between neurosymbolic approaches and learned world models.

Following Anthropic's research output to design rigorous experiments measuring genuine capabilities.