Machine Learning Engineer & AI Researcher
Based in Stockholm. Specializing in teleological, EU-first AI development
Artificial Intelligence through classical philosophy
Engineering is best done when there's a purpose to it.
I've got an MSc in Machine Learning from KTH Royal Institute of Technology in Stockholm. My master's thesis implements Aristotelian virtue ethics into Large Language Models. It's great that models are becoming good at empirical sciences, but due to their mass consumer-base, there's desperate need for models to be good, in the philosophical sense.
I want engineering to be teleological. We ought not innovate for innovation's sake, but for the purposeful advancing of the human condition, so that families and individuals can flourish.
3+
years of startup and AI consulting experience.
I apply my teleological engineering view to everything I build.
Research and Production
From academic research to applied AI engineering.
Machine Learning Research
- MSc Thesis: Hylomorphic AI with Representation Engineering
- ResNet18 with LNR on Imbalanced Music Datasets
- PyTorch, CNNs and diffusion models
- Evaluation and benchmark metrics
Applied AI & Engineering
- Co-founded AI consultancy company early 2023
- Deployed systems with 10k+ chats
- Built RAG and agentic pipelines for clients
- Designed analytics in PostgreSQL and MySQL
- Deployed on Azure and GCP
ML is a field where both continuous research and knowing how to ship code is mandatory
European AI
The US chases commercial hyper-scale and China prioritizes state-controlled systems.
Europemustbuildasovereignalternative,purposefullyintegratedwithinsociety,withafocuson thehumanperson
MSc Thesis: Hylomorphic AI
Virtue-based ethical alignment for LLMs using Representation Engineering (RepE) and Latent Adversarial Training (LAT). Implemented two post-training steering methods to ethically ground LLM behavior within a certain worldview as an alternative to RLHF.
Servai: Applied LLM Systems
Co-Founder and AI Consultant. Built LLM-based systems utilizing OpenAI, Gemini, and Azure AI Hub. Designed complete data analysis pipelines in PostgreSQL/MySQL and deployed scalable agents handling 10k+ chats.
Tellintel: Enterprise Innovation
Co-Founder at Tellintel (KTH Innovation Launch Batch 22). Built a platform to systemize innovation for enterprises. Developed Socratic AI alignment architectures that prompt the user, ensuring AI augments rather than outsources human creativity.
MSc Thesis: Hylomorphic AI
Virtue-based ethical alignment for LLMs using Representation Engineering (RepE) and Latent Adversarial Training (LAT). Implemented two post-training steering methods to ethically ground LLM behavior within a certain worldview as an alternative to RLHF.
Servai: Applied LLM Systems
Co-Founder and AI Consultant. Built LLM-based systems utilizing OpenAI, Gemini, and Azure AI Hub. Designed complete data analysis pipelines in PostgreSQL/MySQL and deployed scalable agents handling 10k+ chats.
Tellintel: Enterprise Innovation
Co-Founder at Tellintel (KTH Innovation Launch Batch 22). Built a platform to systemize innovation for enterprises. Developed Socratic AI alignment architectures that prompt the user, ensuring AI augments rather than outsources human creativity.
Stockholm School of Economics — AGI Keynote (invited by SSE Debate Club)
Matei Cananau
Machine Learning Research Engineer
Production-ready Machine Learning.
My KTH research provides deep theoretical foundations in ML, while my experience co-founding startups ensures I can translate complex AI models into robust, scalable, and intuitive enterprise applications.
What I bring to the table
Theoretical knowledge with software engineering skills.
Engineering & Infrastructure
- Architecting highly available, secure LLM pipelines
- Full-stack development (React, Node, PostgreSQL)
- Deploying and optimizing open-weights models (vLLM)
- Designing scalable, tested cloud infrastructure
Research & ML Theory
- Advanced deep learning and representation engineering
- Rigorous empirical evaluation and KPI tracking
- Post-training alignment (LoRA, DPO, LAT)
- Turning theoretical literature into applied solutions
Everything you need to know.
My core ML stack revolves around PyTorch, transformers, and custom inferencing pipelines (vLLM). For full-stack applications, I specialize in React/Next.js, Node.js, and PostgreSQL, deployed across cloud infrastructure like Azure and GCP.
A lot of ML research papers are abstract and unable to be put into practice. My KTH education provides the theoretical understanding, but my startup experience taught me to prioritize speed, reliability and intuitive user experiences.
At Servai, I architected and deployed autonomous LLM agents handling over 10,000 concurrent interactions, backed by PostgreSQL data pipelines and enterprise-grade security. I translated the wants and needs of the clients into AI solutions.
I am currently based in Stockholm, Sweden. I'm highly open to remote opportunities, but I'm not in a situation where I can relocate.
I speak Swedish, English and Romanian fluently.