AGR
Menu

About / AI engineer, Amsterdam

about me

I'm Andres, an AI engineer in Amsterdam. Getting an AI demo working is easy now; making it hold up in production is the hard part, and that's where I focus: evaluation and regression testing, cost control, observability, and security, from guardrails to prompt-injection defense. Right now that means Cabyn.ai at Boundless Digital, an AI-native SEO platform I took from a single-server MVP to multi-tenant AWS infrastructure as code.

With this blog I want to teach people how to build AI systems that are safe and reliable. These days a million influencers teach the same things: how to automate tasks, how to use Claude or ChatGPT, how to generate animations. Almost no one teaches the kind of engineering it takes to make a product work at production level, so that is what I write about. It is also a way to hold myself accountable to keep learning in my free time.

View projects Get in touch

01 / Experience

9 entries
Sep 2025 – NowNowAI EngineerBoundless DigitalLeading Cabyn.ai, an AI-native SEO platform, plus AI consulting and training.
  • AI Technical Solution Engineer Lead, owning the full lifecycle of solutions across domains.
  • Product ideation and development of Cabyn.ai, an AI-native SEO analysis and content creation system for a digital marketing agency's client base.
  • Migrated its MVP from a single server to multi-AZ, autoscaling, multi-tenant AWS infrastructure managed entirely as code: networking, IAM, zero-egress compute and delivery layers across 4 environments, with a zero-destroy migration guarantee and CI-driven plan-on-PR review.
  • Established CI/CD best practices and automated unit and smoke tests for production reliability and uptime.
  • Scaled from a couple of users to over a dozen, with 600%+ year-over-year revenue growth.
  • Led design and implementation of custom AI automations for clients, from agentic harnesses (Claude Code) to custom-built solutions (LangGraph).
  • Trained internal consultants and external clients on automation platforms (Claude Code, Cowork) to drive hands-on adoption.
  • AWS
  • Terraform
  • CI/CD
  • LangGraph
  • Claude Code
Jan 2025 – Sep 2025AI EngineerCustomAIBuilt VisionPay, an invoice OCR tool that was later acquired.
  • Designed, built and deployed VisionPay, an OCR tool for invoice reconciliation and the startup's first revenue-generating product, since sold and integrated into a private company.
  • VisionPay reached 99.8% accuracy and cut monthly processing-error costs by $1.1M, an extrapolated $13M+ per year.
  • Built a multimodal RAG pipeline for thousand-page documents, combining vector databases with a knowledge graph.
  • Built an LLM-based agentic framework that scrapes API reference sites and generates code for each API call it finds.
  • LangChain
  • FastAPI
  • AWS
  • Amazon Bedrock
  • RAG
Jul 2024 – Sep 2025AI EngineerVrije Universiteit AmsterdamEvolutionary robotics framework, MLOps and LLM-based swarm control research.
  • Upgraded and maintained Revolve2, an evolutionary robotics framework that co-evolves the bodies and brains of modular robots, used by the Collective Intelligence research group.
  • Ran the group's MLOps: MLflow tracking servers on AWS (EC2, S3, PostgreSQL) and CI/CD pipelines for testing and deploying versioned models.
  • Continued research on natural-language control of multi-agent systems and robot swarms using LLMs.
  • Python
  • MLflow
  • AWS
  • CI/CD
  • LLMs
Dec 2023 – Jul 2024Research AssistantVrije Universiteit AmsterdamSwarm robotics and collective intelligence research, plus thesis supervision.
  • Research in swarm robotics, collective intelligence, exploration and reinforcement learning at the robotics lab.
  • Guided bachelor and master students through their thesis research.
  • Worked with a robotics startup to design and test product use cases, reporting bugs and proposing features.
  • Swarm Robotics
  • Reinforcement Learning
  • UAVs
Oct 2023 – Jun 2024Teaching AssistantVrije Universiteit AmsterdamTA for Multi-Agent Systems, Learning Machines and Collective Intelligence.
  • Project Collective Intelligence: created assignments, gave lectures introducing them, and ran practicals on flocking, aggregation and predator/prey scenarios. Also helped structure the course.
  • Learning Machines: reviewed and improved the course codebase, ran practicals and graded assignments on a wheeled robot (simulation and hardware) using evolutionary computing and reinforcement learning.
  • Multi-Agent Systems: ran practical sessions, graded assignments and guided students through multi-agent tasks in Prolog and MARBEL.
  • Swarm Intelligence
  • Evolutionary Computing
  • Reinforcement Learning
  • Prolog
Sep 2022 – Jan 2025Chief Operating OfficerMax Level StudiosCOO after Max Level acquired my studio. Helped raise $250K.
  • Joined as COO after Max Level acquired my previous venture, Middleverse Studios.
  • Helped strategise a successful $250K community fundraise.
  • Managed relationships and partnerships with industry leaders such as Immutable.
Aug 2021 – Sep 2022Founder & CEOMiddleverse StudiosFounded a blockchain game studio, raised $150K pre-seed, negotiated its sale.
  • Founded and led Middleverse Studios, a blockchain game studio.
  • Built business and investor relationships to raise $150K in a pre-seed round.
  • Assembled a team of industry veterans from studios such as Ubisoft and Marvel.
  • Negotiated the studio's acquisition by Max Level Studios.
Mar 2019 – Aug 2021Blockchain ResearcherInBestAnalysed blockchain projects and brought investment proposals to the trading team.
  • In-house blockchain researcher providing consulting services.
  • Analysed promising blockchain projects and brought investment proposals to the trading team.
Oct 2014 – Feb 2019Blockchain Investment & MiningSelf-employed · BogotáBuilt GPU mining rigs for clients and advised on crypto portfolios.
  • Built GPU-based cryptocurrency mining rigs for private clients at different scales.
  • Consulted for clients diversifying their portfolios into crypto-based projects.
  • Researched and invested in blockchain projects, and ran my own GPU mining operation.

02 / Education

2 entries
2023 – 2025MSc Artificial IntelligenceVrije Universiteit AmsterdamGraduated 8.1/10. Thesis on controlling robot swarms through natural language.
  • Final grade 8.1/10 (3.9/4 GPA).
  • Thesis (9.1/10): MinionsLLM, a task-adaptive framework for the control of multi-agent systems through natural language.
  • Designed a method to generate synthetic fine-tuning data from a formal grammar for behaviour trees, reducing bias and hallucinations in multi-agent control tasks.
  • Coursework projects spanning NLP, recommender systems, computer vision for drone control and time-series prediction.
  • LLMs
  • Multi-Agent Systems
  • Fine-tuning
  • PyTorch
2019 – 2023BSc Artificial IntelligenceVrije Universiteit AmsterdamGraduated 7.6/10. Thesis on multi-agent RL for collective game agents.
  • Final grade 7.6/10 (3.7/4 GPA).
  • Thesis (8.5/10): EMiX, a multi-agent reinforcement learning model for engaging behaviour of collective game agents.
  • Reinforcement Learning
  • Multi-Agent Systems
  • Python