irisdania.dev

Hi, I'm Iris

Phd Student in Embedded System Security.

See my work Get in touch
Munich, Germany · Open to collaborations
Iris Dania Jimenez

About

My research intrest lies at the intersection of AI, cryptography and hardware. My background is in Machine Learning and Statistics. I’ve worked on explainable AI, computer vision, and NLP model merging. In my free time I like to read. A lot.

Education

Ludwig Maximilian University of Munich

Master’s in Statistics & Data Science · Apr 2023 – Oct 2025

Minor in Machine Learning.

University of Milan‑Bicocca

BSc in Statistics & Information Management · Oct 2019 – Mar 2023

Coursework: Python, SQL/NoSQL, R, SAS; supervised & unsupervised learning; statistical modelling; demography; medical/health statistics; Tableau/Orange/RapidMiner; parallel DBs & queries.

Experience

PhD Student · Embedded System Security Group & AI @ UniBw Munich

Nov 2025 – Present · Munich

  • AI for Side Channel Analysis
  • AI for Hardware Reverse Engineering
  • Skills: Python, Cluster configuration, Git

Research Assistant · Chair of Computer Vision & AI @ TUM

Nov 2024 – Sept 2025 · Munich

  • Collaborated with a PhD candidate on Computer Vision research.
  • Implemented algorithms and simulation models.
  • Conducted literature research in XAI and CV.
  • Skills: Python, Cluster configuration, Git

Research Assistant · Bavarian AI Chair for Mathematical Foundations of AI

Jun 2023 – Feb 2025

  • Advanced research in explainable AI (XAI).
  • Implemented algorithms and simulation models.
  • Comprehensive literature research in XAI and CV.
  • Skills: Python, Cluster configuration, Git

Machine Learning Intern · EOMAP GmbH (Germany)

Oct 2022 – Dec 2022

  • Analyzed satellite images using unsupervised ML techniques.
  • Developed a classification algorithm to detect cloud shadows.
  • Skills: Python, QGIS, AWS, k-means

Selected Projects

VI-MIDAS: Variational inference for microbiome survey data

Extended VI-MIDAS for microbial abundance modeling with latent ecological factors; applied to a new marine dataset. Built a variant enabling unified latent representations of environmental drivers. Project page

Model Merging for NLP

Combined TinyBERT and BERT-Base to balance accuracy and efficiency. Evaluated on IMDb & SNLI (GLUE), comparing accuracy and FLOPs against standard fine-tuning. Project page

Publications & Thesis

Languages

Contact

Reach out for collaborations, research, or ML projects.