Amirhoshang Hoseinpour Dehkordi

Proving what
machines know.learn.believe.share.

I build the mathematics of trust in software — bridging formal verification and epistemic logic to reason rigorously about what intelligent systems know.

View the work Read the thesis ↗

Ph.D., IPM · Teaching Faculty, Sharif University of Technology · Kish Island

order chaos
fig. 01 — a field of knowing agents: structure ⇄ entropy
scroll
00 / Profile

I am a computer scientist and teaching faculty member at the Sharif University of Technology, on Kish Island. I earned my Ph.D. from the Institute for Research in Fundamental Sciences (IPM), where my dissertation asked a precise question: when an intelligent system claims to know something — can we prove it?

My research sits where mathematical logic meets machine learning. To reason about the knowledge classifiers and neural networks hold, I introduced MASKS and LTPAL. Alongside the theory I have spent fifteen years as a working engineer, and co-founded PersMed.

15+ years of software 9 peer-reviewed works 3 universities 2 logics introduced
01 / Research

Proving what machines know.

primary focus

Formal Verification of Multi-Agent Learning Systems

MASKS treats a collection of classifiers as a society of knowing agents. Instead of testing on averages, it performs point-wise verification: for a single input, a property is checked against the shared knowledge of the classifiers. When the agents agree, the input verifies the property for the system. LTPAL extends Public Announcement Logic with Linear Temporal Logic, reaching properties of entire data streams.

MASKS · LTPAL · PAL · Model Checking Dissertation ↗

Epistemic Logic for Neural Networks

A network's layers, read as agents that acquire and announce knowledge — a language for what a model has learned at each stage.

FSEN 2019

Epidemic Data & Statistics

Uncovering the real statistics behind COVID-19 testing and case-fatality data, separated from the artefacts of collection.

J. Medical Virology · 2020–21

Distributed Knowledge & Verification

Meet MASKS — a uniform framework combining what a group of classifiers knows with information from other sources.

Soft Computing · 2025
02 / Publications

Selected works.

  1. 2025

    Meet MASKS: Integrating Distributed Knowledge and Verification for Multi-Agent Systems

    Soft Computing
  2. 2023

    Linear Temporal Public Announcement Logic: reasoning about the knowledge of multi-classifiers

    Bulletin of the Iranian Mathematical Society
  3. 2021

    What does COVID-19 Testing Really Say? The Real Statistics Concealed Behind the Accessible Data

    Journal of Medical Virology · doi:10.1002/jmv.27173
  4. 2020

    Understanding Epidemic Data and Statistics: A case study of COVID-19

    Journal of Medical Virology · doi:10.1002/jmv.25885
  5. 2019

    An Epistemic Logical View of Deep Neural Networks

    Fundamentals of Software Engineering (FSEN)
  6. 2016

    Model Checking Multi-Agent Systems

    MSc thesis · University of Tehran · GPA 3.9, rank 2
  7. 2013

    Assessing New Conditions for the Secretary Problem using Multi-Agent Systems

    IEEE — 13th Iranian Conference on Fuzzy Systems

reviewer — BMJ Open · Journal of Medical Virology · Neural Computing and Applications · Springer Nature Trusted Reviewer

03 / Experience

From theory to production.

2024 —

Teaching Faculty

Sharif University of Technology, International Campus

2020 —

Co-Founder

PersMed

2024–25

Teaching Faculty

University of Tehran, International Campus

2019–21

Technical Consultant

University of Tehran, International Campus

2018–19

Strategy Manager & Data Scientist

Shaghayegh Co. — Python

2016–18

Senior Game & Back-end Developer

XroboX Co., UT Science & Tech Park — Python, C++

2013–14

Senior C++ Developer

XroboX Co. — Robotics, simulation & modeling

2009–11

Software & Web Developer

SafaMed & Goharhonar — C#, JS, SQL

2007–09

Software Developer

Negaresh Rayaneh Poya — Image processing, C++

languagesPython · C++ · Java · C# · JavaScript
learningTensorFlow · PyTorch · Scikit-learn · NumPy · Repast · NetLogo
dataPostgreSQL · MySQL · SQL Server · Oracle
systemsOpenCV · Halcon · AtmelStudio · Unreal Engine · Cocos2D-X
04 / Teaching & Honors

Courses taught

  • SharifFundamentals of Programming
  • SharifData Structures
  • SharifAdvanced Programming
  • SharifDiscrete Structures
  • UTFormal Methods
  • UTStatistics & Probability
  • UTGeneral Mathematics 2

Recognition

  • 2nd rank, Iran CS Ph.D. nationwide entrance exam — 2014
  • Bronze medal, National Informatics Olympiad — 2003 & 2004
  • Diploma of Honor, Kharazmi National Festival of Innovation — 2002
  • IPM Foundation graduate scholarship — 2016–2021
  • National Elite Foundation fellowships — 2007–2015
05 / Students

Bachelor theses supervised.

  1. i. Mohammad Mohammadkhani
  2. ii. Nika Sadidi
  3. iii. Parsa Molayi
  4. iv. Hosein Dehghan
06 / Contact

Let's reason together.

Open to research collaboration, talks on formal methods and trustworthy AI, and advising.

Write to me Curriculum Vitae ↗