The person behind the work / About
About
From control engineering and tensor methods to building practical AI tools, I like turning complex ideas into things people can use.
View my LinkedIn
I’m Pierre-Antoine Denarié. I like turning difficult ideas into things people can test, see, and use. My route into AI started with mechanical engineering and control theory, then moved through tensor methods and practical data systems.
Engineering roots
I studied Mechanical Engineering at TU Delft before completing a master’s in Systems and Control, graduating cum laude in 2022. The combination shaped how I work: model the system, understand its limits, and make the result useful outside a notebook.
My master’s thesis explored independent component analysis through implicit cumulant tensor decomposition. Alongside it, I worked on tensor train decomposition for neural network weight compression and tensor-based ICA for fMRI neural imaging. Those projects brought together the areas I still return to: signal processing, optimization, state-space models, and machine learning.
Work in practice
I joined Deloitte in 2023 as a Business Analyst, later became a Consultant, and now work as an AI Consultant in AI & Data. Earlier work included building full-stack data migration tooling, working with procurement processes in SAP transformations, and improving data validation tools. I’m interested in the point where strong technical ideas meet the constraints of a real organization.
Before Deloitte, I worked on an HVAC modeling method at Damen and served on TU Delft’s Faculty Student Council. I also spent a year on the board of the mechanical engineering study association Gezelschap Leeghwater, where I learned a great deal about working with people and making systems run better.
Things I’ve built
- The Cocktail Party Problem turns my thesis work into an interactive audio demo. You can mix signals, tune the ICA pipeline, and hear what separation actually does.
- Structured model compression of RF-DETR explores tensor train and CP factorization, plus low-precision core storage, for an image segmentation model.
- Tensor Decomposition Study is a hands-on exploration of Tucker1, TT-SVD, and TTm-SVD methods.
- Bus times estimation case study is another project listed in my portfolio.
These and my other projects are listed on LinkedIn. I’m also exploring local AI tools and their practical capabilities, costs, and tradeoffs; this blog is where I share those experiments and the reasoning behind them.
Credentials and beyond
My certifications on LinkedIn include:
- SAP Certified Associate — Data Engineer — SAP HANA (2025)
- SAP Certified Associate — SAP S/4HANA Cloud Private Edition, Sourcing and Procurement (2025)
- Customer Value in Pricing Strategy and Cost and Economics in Pricing Strategy, Coursera (2019)
- Cambridge Certificate of Proficiency in English C2 (2014)
Karate has been part of my life since childhood. I hold a second dan black belt, and the discipline of practicing hard things over time has stayed with me well beyond the dojo.
For the full professional timeline, connect with me on LinkedIn.