Matei Cananau
AI/ML Engineer and Data Scientist
Stockholm, Sweden • KTH Royal Institute of Technology (MSc Machine Learning, BSc Media Technology, Civilingenjör)
I am a Machine Learning Engineer and Systems Architect specializing in bridging deep theoretical machine learning (representation engineering, activation steering, latent space geometry) with robust production engineering (agentic workflows, SQL pipelines, enterprise API orchestration).
Having co-founded an AI consultancy deploying systems with 10,000+ live customer interactions, built enterprise innovation platforms accepted into KTH Innovation Launch Batch 22 and guest lectured AI alignment at KTH, I've got a great mix of engineering, scientific computing and teleological, human-centered philosophy.
Featured Research & Papers
Theoretical breakthroughs in neural representation geometry and virtue alignment.
Hylomorphic AI: Virtue-Based Ethical Alignment
A white-box alternative to standard RLHF. Reframing alignment as the active imposition of structural form upon internal activations using Representation Engineering (RepE) and Linear Artificial Tomography (LAT) on Google Gemma 4.
Label Noise Regularization for Audio Datasets
Implemented Label Noise Regularization (LNR) to shift decision boundaries in highly imbalanced audio classification datasets. Trained ResNet-18 architectures on Mel-Frequency Cepstral Coefficients (MFCC), earning Grade A at KTH.
Digital Mindfulness & Cognitive Systems
Investigated cognitive relationships between intentional digital breaks, focus retention and psychological well-being. Selected by faculty as benchmark reading for subsequent media technology students at KTH.
Concept Direction & Vector Steering
Select a virtue below to examine its extracted concept direction from Google Gemma activation space:
Prudence
Ethical Definition: Practical wisdom discerning context, appropriate boundaries and strategic foresight in generative responses.
Activation Steering Mechanism: Stateful Subspace Projection ensuring reasoning chains evaluate unintended secondary consequences.
Engineering & Leadership
Co-Founder & AI Engineer — Tellintel
Stockholm, Sweden • 2025 – Present
- Selected for KTH’s premier deep-tech incubation program to build an enterprise platform systematizing innovation workflows.
- Engineered a proprietary Socratic AI interaction architecture that guides enterprise users through structured problem-definition and prompt-reasoning flows rather than outsourcing strategic thinking to models.
- Developed scalable full-stack system architecture integrating real-time frontend interfaces with backend LLM orchestration pipelines.
Co-Founder & AI Consultant — Servai
Stockholm, Sweden • 2023 – 2026
- Founded and operated an AI consultancy designing, building and deploying custom GenAI agents and automated support systems for Swedish commercial clients.
- Engineered end-to-end data pipelines in Python with MySQL and PostgreSQL databases hosted on Azure AI Hub and Google Cloud Platform (GCP), ensuring strict compliance with GDPR and the EU AI Act.
- Deployed production AI agent systems supporting over 10,000 live client chats, tracking model performance metrics and delivering technical architecture reviews directly to executive stakeholders.
Academic Leadership & Teaching — KTH Royal Institute of Technology
Stockholm, Sweden • 2021 – 2026
- Guest Lecturer in AI Alignment (2026): Designed and taught an AGI course module covering AI safety, model alignment and technical evaluation frameworks to senior engineering students.
- Teaching Assistant (TA) in Programming & Scientific Computing (2022 – 2023): Taught Python, numerical algorithms and problem-solving methodologies for undergraduate engineering courses.
- Academic Excellence: Completed 303.5 ECTS credits with straight A grades in Deep Learning Advanced, Foundations of Machine Learning, Applied Computer Science, Programming & Scientific Computing and Information Visualization.
Essays & Philosophical Writing
Essays from the Digital Monastery exploring the intersection of AI architecture, semiotics and philosophy.
Introducing Artificial Intelligence to Metaphysics
Bridging deep learning with classical ontology, analyzing how neural representations relate to form, substance and computation beyond mere statistical pattern matching.
The Lack of Soul in AI Systems
An inquiry into intentionality, consciousness and why next-token prediction and functional output fundamentally lack the qualitative, teleological essence of the human soul.
EU AI Act and Europe’s AI Philosophy
Thoughts on the regulatory framework and safety mandates of the EU AI Act from the perspective of European technological sovereignty and philosophical traditions.
Technical Expertise
ML & AI
- • PyTorch & Transformers
- • Representation Eng.
- • Linear Artificial Tomo.
- • Mechanistic Interp.
- • Activation Steering
- • RAG & Agents
- • Prompt Evaluation
- • Diffusion & CNNs
Data & Backend
- • Python (NumPy, Pandas)
- • SQL (PostgreSQL, MySQL)
- • REST APIs & FastAPI
- • Docker Containers
- • Linux / Bash
- • Git / GitHub
- • Data Pipelines
Cloud & AI Law
- • Microsoft Azure AI
- • Google Cloud (GCP)
- • MLOps Deployment
- • EU AI Act
- • GDPR Compliance
- • Annex IV Technical
Frontend & Delivery
- • JavaScript / TypeScript
- • React & Next.js
- • HTML5 / CSS3
- • Visualization
- • Socratic Interface
- • Technical Lecturing
- • Executive Reviews
Questions & Inquiries
Get in Touch
I am open to Machine Learning Engineering, AI Research roles and Data Science positions.
Stockholm, Sweden • 2026
Matei Cananau — ML/AI Engineer & Data Scientist