Skip to content
Mehdi Dehghani Firoozabadi

About

I am a computer scientist based in Tehran. After a software-engineering degree at Yazd University and years building production systems — CRM, Persian OCR, access control — I returned to research on generative models. The through-line is the same: compact methods that do something precise.

Download resume
Mehdi Dehghani Firoozabadi

Education

  • M.Sc. in Computer Science · Data Mining

    2020 — present

    Tarbiat Modares University, Tehran

    • Thesis: "An Improved Generative Model Based on Diffusion"

  • B.Sc. in Computer Engineering · Software

    2002 — 2005

    Yazd University, Yazd

    • Graduated in seven semesters, second rank
    • Thesis: "Design and Implementation of a Real Estate Web System"

Experience

  1. Startup MMMF

    2024 — present

    Tehran, Iran

    • Developed a time-series forecasting framework for multivariate data using deep learning models.
    • Built data pipelines for preprocessing, transformation, and feeding datasets into the forecasting framework.
    • Applied feature-importance methods to identify and select informative variables for modeling.

  2. Freelance instructor

    2020 — present

    Tehran, Iran

    • Teaching Python programming, machine learning, and deep learning.

  3. Rahian Software Development

    2015 — 2019

    Tehran, Iran

    • Collaborated on an intelligent gate-barrier system for parking-lot access control.

  4. Sepanta Robotic Research Foundation (SRRF)

    2014 — 2015

    Tehran, Iran

    • Collaborated on an OCR system for Persian scripts.

  5. Mehregan Danesh Pazhooh

    2009 — 2014

    Tehran, Iran

    • Collaborated on in-house CRM software.

Research interests

  • Interpretable deep learning, with emphasis on compact, parameter-efficient architectures
  • Deep generative models, including diffusion models and rectified flow
  • Inverse problems using score-based diffusion models
  • Exploration of latent space through dimensionality reduction
  • Time series forecasting
  • The influence of consciousness on the generation of random numbers

Skills

Languages
Python, C++, Delphi
Machine learning
PyTorch, NumPy, Pandas, SciPy, scikit-learn, OpenCV
Experimentation
Hydra, Optuna, MLflow, Ray, Captum, TensorBoard
Other
Git, Linux, LaTeX

Honors

Exams

  • Ranked 25th nationally in the M.Sc. Computer Engineering entrance exam (2006)
  • Ranked 37th nationally in the M.Sc. Computer Science entrance exam (2006)

Software & competitions

  • Provincial winner, 4th Kharazmi Young Festival, for real-time voice conversion software (2002)
  • Official software registration at the High Council of Informatics (2006)
  • Winner of the Sharif Business Plan Competition for mobile text-to-speech software (2007)

Innovation festivals

  • Selected, 2nd Innovation Festival, University of Tehran — cloud-based attendance tracking (2017)
  • Selected, 3rd Innovation Festival, University of Tehran — book provision and exchange (2018)
  • Selected, IdeaTap, Qom Science and Technology Park — simulator for sensitive environments (2019)
  • Selected, Innovation Festival, University of Tehran — VR/AR simulator for sensitive environments (2020)

Academic notes

Course presentations have included pitch estimation with fuzzy methods, rotation-invariant face recognition, voice transformation, time-varying spectrum estimation, virtual reality, evolutionary multi-objective optimization, tensor methods in vision, discrete contrastive diffusion, contrastive clustering, and denoising diffusion models.