Loading experience...

0:00

Jay | Ludovico Einaudi

Ali Vaezi

Hello, I'm

Ali Vaezi

Reinforcement Learning Researcher Optimization Engineer AI Research Engineer MSc @ Politecnico di Torino

Building intelligent systems at the intersection of RL, optimization, and real-world decision-making.

About Me

I am a final-semester MSc student at Politecnico di Torino specializing in Reinforcement Learning, Applied Machine Learning, and Mathematical Optimization, with a BSc in Industrial Engineering from Azad University (SRBIAU) (GPA 3.69/4.0) and 5 years of professional experience as a Business Data Analyst at Parva Co.

My research builds intelligent systems for real-world decision-making under uncertainty, from supplier order allocation with PPO to generalization studies in sim-to-sim transfer. My MSc thesis at TU Eindhoven develops data-driven remaining-useful-life estimation coupled with an economic decision model for maintenance timing. I carried out a research internship at TU Eindhoven (Mar–Jul 2026) and am joining Blue Engineering as an AI Engineer Intern. I am a fast learner who ramps quickly into new technical domains. I have published in the Journal of Cleaner Production (IF 11, 52 citations) and have a paper under review at the International Journal of Production Economics.

0+ Citations
0 Publications
0 GRE · Q 169
0 Projects
English C1/C2 Persian Native Italian A2

Experience & Education

AI Engineer Intern Incoming

Jul 2026 – Jan 2027

Blue Engineering Srl · Turin, Italy

Applying machine learning and AI methods to engineering problems in a six-month industrial placement.

Research Intern (Curricular)

Mar 2026 – Jul 2026

Eindhoven University of Technology (TU/e) · Netherlands

MSc thesis research on remaining-useful-life estimation and economic intervention timing under partial observability.

Machine Learning Researcher Part-time

Mar 2026 – Present

MALTO Research Team, Politecnico di Torino · Turin, Italy

Deep learning research and competitive ML benchmarking (Macro F1 0.959).

Teaching Assistant, AI & Machine Learning

Sep 2025 – Present

Politecnico di Torino · Turin, Italy

Support graduate coursework on ML, optimization, and applied data analysis (Prof. Danilo Giordano).

Teaching Assistant, Python & Operations Research

2020 – 2023

Azad University (SRBIAU) · Tehran, Iran

Taught programming and quantitative methods to undergraduate cohorts.

MSc Thesis

Eindhoven University of Technology (TU/e)

MSc Thesis Research

Eindhoven, Netherlands

QS #152 Sep 2026 – Feb 2027

Data-driven remaining-useful-life estimation coupled with an economic decision model for maintenance timing under uncertainty. Supervisors: Prof. Mehrdad Mohammadi (TU/e), Prof. Lorenzo Bottaccioli (PoliTO).

MSc

Politecnico di Torino

Digital Skills for Sustainable Societal Transitions

Turin, Italy

QS #207 2024 – 2027 101/110

Reinforcement Learning, Mathematical Optimization, Deep Learning, Probabilistic Modelling, IoT. Expected graduation in the spring 2027 session.

BSc

Azad University (SRBIAU)

Industrial Engineering

Tehran, Iran

QS #151–175 2019 – 2023 GPA 3.69/4.0

Operations research, industrial optimization, quality management. BSc thesis published in Journal of Cleaner Production (IF 11, 52+ citations) on blockchain-integrated sustainable supplier selection.

Technical Skills

Core competencies across research and engineering

Programming

8 tools
PythonSQLR MATLABCGitLinuxLaTeX

ML & Deep Learning

6 tools
PyTorchTensorFlowScikit-learn NumPyPandasOpenCV

Reinforcement Learning

7 tools
PPODQNSAC REINFORCEActor-CriticGymnasiumMuJoCo

Optimization

7 tools
GurobiCPLEXGAMS MILPBWMTOPSISFuzzy MCDM

Data & Analytics

6 tools
PandasMatplotlibQGIS Power BISPSSExcel

IoT & DevOps

7 tools
MQTTRaspberry PiDocker REST APIsFastAPIReactThree.js

My Projects

Research projects and open-source contributions in AI, RL, and optimization.

MSc Thesis

RUL-Driven Maintenance & Remanufacturing Decisions

Researchers: A. Vaezi

Affiliation: Eindhoven University of Technology (TU/e) · Industrial Engineering & Innovation Sciences

Status: MSc Thesis | In Progress 2026

Keywords: Remaining Useful Life, Probabilistic Degradation, Economic Decision Model, Sensor Data, Circular Remanufacturing

Data-driven remaining-useful-life estimation from run-to-failure sensor data, coupled with an economic decision model that times maintenance and selects remanufacturing pathways (reuse, recycle, conventional, additive manufacturing) under uncertainty. Supervisors: Prof. Mehrdad Mohammadi (TU/e), Prof. Lorenzo Bottaccioli (PoliTO).

View on GitHub
CLiC-it 2026

xAI Hallucination Benchmark

Researchers: A. Vaezi

Affiliation: Politecnico di Torino · DIST

Venue: CLiC-it 2026 | In Preparation

Keywords: Vision-Language Models, Hallucination, Attention Rollout, Grad-CAM, Logit-Lens, AGTR

Cross-domain benchmark applying three explainable-AI methods to diagnose why vision-language models (LLaVA variants, LLaVA-Med, Qwen-VL, BLIP-2) hallucinate across medical (PathVQA, VQA-RAD, HALT-MedVQA) and industrial (MVTec AD, MMAD) inputs. Introduces AGTR (Attention-Guided Token Rescoring), a training-free mitigation algorithm yielding POPE F1 Δ = +0.081.

View on GitHub
Under Review

Tehran NL-SUE

Researchers: A. Vaezi, S. Alitanlo, A. Naghedi Ahmadi, Prof. C. Pronello

Affiliation: Politecnico di Torino

Venue: Under Review

Keywords: Nested Logit, Stochastic User Equilibrium, Multimodal Transport, Value of Time, MATLAB

Nested-Logit Stochastic User Equilibrium model for multimodal transportation in Tehran, jointly capturing perception variability (stochastic equilibrium), hierarchical substitution among correlated modes (nested logit), and income-based heterogeneity in value of time under strategic interaction. MATLAB simulation with reproducibility scripts.

View on GitHub
Research

North Star Metric Selection (BWM-TOPSIS)

Researchers: A. Vaezi, A. Karimian, Prof. G. Santoro

Affiliation: Politecnico di Torino & University of Turin

Status: Manuscript in Preparation

Keywords: BWM, TOPSIS, MCDM, Startup Metrics, Business Model Archetypes

Context-aware MCDM framework that operationalizes how early-stage startups should select their North Star Metric given business-model archetype (SaaS/Subscription, Marketplace/Transactional, B2B/Enterprise). Best-Worst Method weights eight validated criteria; TOPSIS ranks candidate metrics per archetype, showing NSM selection is not context-neutral.

View on GitHub
Full-Stack

Finance-AI Trading System

Researchers: A. Vaezi, S. Shahali

Affiliation: Personal Research Project

Keywords: RL Trading, Stable-Baselines3, Gymnasium, FastAPI, ClickHouse, React, Docker

Production algorithmic trading system with an RL agent layer (Stable-Baselines3 + Gymnasium), XGBoost signal models, and a full-stack serving pipeline: FastAPI + WebSockets backend, ClickHouse time-series storage, APScheduler jobs, and a React dashboard. Containerized with Docker.

View on GitHub
ML + Security

IIS Lens

Researchers: A. Vaezi, Vali

Affiliation: Politecnico di Torino · 2026

Keywords: Isolation Forest, Anomaly Detection, FastAPI, IIS Logs, Pydantic, TDD

Windows-native IIS log analytics that turns raw W3C logs into plain-English security and performance verdicts in seconds. Isolation Forest plus IIS-specific rules flag brute-force, scanners, and DDoS bursts, with prescriptive next-step actions. FastAPI backend, Pydantic schemas, TDD-first pytest suite, and CSV/PDF export.

View on GitHub
Open Source

pygma

Researchers: A. Vaezi

Venue: PyPI · GitHub · 2026

Keywords: General Morphological Analysis, Cross-Consistency Assessment, Decision Support

First open-source Python implementation of Fritz Zwicky's General Morphological Analysis (GMA). Defines a morphological field, applies a Cross-Consistency Assessment, and reduces the solution space to internally consistent scenario archetypes — bringing a method previously locked in proprietary software into the open-source ecosystem.

View on PyPI View on GitHub
Live App

AliFit Tracker

Developer: A. Vaezi

Stack: Next.js 16 · TypeScript · Supabase · Vercel

Keywords: Full-Stack, Progressive Overload, Nutrition Tracking, Supabase Auth

Body-recomposition web app: 6-day Push/Pull/Legs training tracker with progressive-overload charts and deload management, a calorie/macro nutrition logger, a standalone BMR/TDEE calculator (Mifflin-St Jeor), and a progress dashboard with ETA-to-goal. Next.js 16 App Router, TypeScript, Supabase Postgres + auth, deployed on Vercel.

Live Demo View on GitHub
Research Project

SUPRA-PPO

Researchers: A. Vaezi, E. Rabbani, Prof. G. Bruno

Affiliation: Politecnico di Torino

Target: Journal of Cleaner Production | In Preparation

Keywords: Fuzzy BWM, Fuzzy TOPSIS, PPO, Supply Chain, Order Allocation

Two-phase framework combining Fuzzy BWM + Fuzzy TOPSIS with a custom PPO agent for supplier order allocation under stochastic demand, lead-time uncertainty, and systemic disruptions. Gymnasium-based supply chain simulator with three market scenarios (stable, volatile, disrupted).

View on GitHub
Research

Sim-to-Real Locomotion

Researchers: A. Vaezi, Y. Fayyaz, S. Shahali, P. Hashemi Alvar

Affiliation: Politecnico di Torino

Status: Independent research project

Keywords: PPO, REINFORCE, Actor-Critic, MuJoCo, Domain Randomization, ES-CDR

Trained robust locomotion policies (REINFORCE, Actor-Critic, PPO) with domain randomization and curriculum learning (ES-CDR) on a custom MuJoCo Hopper with randomized friction, mass, and joint properties. Benchmarked 6 physics parameters across 3 RL algorithms.

View on GitHub
#1 Leaderboard

WeldFusionNet

Researchers: A. Vaezi, S. Shahali, K. Salimi

Affiliation: Politecnico di Torino · I3P Hackathon 2026

Keywords: Multimodal Fusion, 1D-CNN, MobileNetV3, Focal Loss, Temperature Scaling

Multi-task fusion network for 7-class weld defect classification fusing sensor (1D-CNN), audio (1D-CNN + MFCC), and video (MobileNetV3-Small) branches. 15-step pipeline with focal loss, temperature scaling, and confidence-gated reclassification. F1 = 0.957. Includes Streamlit monitoring dashboard.

View on GitHub
Hackathon

MoveWise

Researchers: A. Vaezi, S. Shahali, K. Salimi, Prof. C. Pronello

Affiliation: Politecnico di Torino · NEXUS 2026 Hackathon

Keywords: DQN, MaaS, Behavioral Nudging, React, Three.js, FastAPI, Docker

Mobility-as-a-Service super-app using DQN with a Habits-Utility-Rationality behavioral model for personalized multimodal route recommendations. A second RL agent selects behavioral nudges to encourage green transport. React 18 + Three.js frontend, FastAPI + PyTorch backend, Docker deployment.

View on GitHub
Research

Turin Micromobility Analytics

Researchers: A. Vaezi, Prof. C. Pronello

Affiliation: Politecnico di Torino · 2024–2025

Keywords: STL Decomposition, Kaplan-Meier, Moran's I, Monte Carlo, Kaggle DOI

Analysis of 2.5M+ shared e-scooter trips in Turin (LIME, VOI, BIRD). STL decomposition, Kaplan-Meier survival analysis, Moran's I spatial autocorrelation, and Monte Carlo simulation (10,000 iterations). Dataset published on Kaggle with DOI.

View on GitHub
IoT

Smart Precision Irrigation

Researchers: A. Vaezi, R. Tossato Silva, L. Deriu, N. Restrepo Lopez

Affiliation: Politecnico di Torino · 2025–2026

Keywords: MQTT, REST APIs, Raspberry Pi, Microservices, Telegram

6 Python microservices for precision agriculture: soil moisture sensing, weather API integration, weather-aware automation scheduling, and real-time Telegram monitoring. Deployed on Raspberry Pi with MQTT protocol and REST APIs.

View on GitHub
Open Source

OpenCLD

Researchers: F.M. Ottaviani, G. Zenezini, A. Ferrari, M. Rebuglio, P. Viero, F. Incerti, A. Vaezi, M.O. Nesanir

Venue: PyPI · GitHub · 2025

Keywords: System Dynamics, Causal Loop Diagrams, Stock-Flow, Time-Series Simulation

Open-source System Dynamics Simulation Library on PyPI. Implements causal loop diagrams, stock-flow modelling, and time-series simulation with clean Python API and comprehensive documentation.

View on PyPI View on GitHub
Live Competition

MALTO: AI Authorship Detection

Researchers: A. Vaezi

Affiliation: Politecnico di Torino · MALTO Recruitment Hackathon 2026

Keywords: DeBERTa-v3, Focal Loss, TF-IDF, Ensemble Stacking, Macro F1, NLP

6-class text classification pipeline distinguishing human-written text from five AI models (DeepSeek, Grok, Claude, Gemini, ChatGPT) under severe 19:1 class imbalance. 8-phase reproducible pipeline combining 46 stylometric features, DeBERTa-v3-base fine-tuned with Focal Loss, and a stacking ensemble. Achieved OOF CV F1 = 0.9591 and public leaderboard F1 = 0.9217.

View on GitHub
Research

NYC Urban Heat Risk Mapping

Researchers: A. Vaezi

Affiliation: Politecnico di Torino · 2025

Venue: Computers, Environment and Urban Systems | Under Review

Keywords: Fuzzy AHP, QGIS, Landsat 8, GIS, Heat Risk Index, Zenodo

Multi-indicator urban heat risk mapping across NYC boroughs (2013–2023). Fuzzy AHP weights multi-criteria indicators; QGIS Graphical Modeler automates geospatial processing; Landsat 8 Level-2 imagery provides Land Surface Temperature and NDVI. Dataset permanently archived on Zenodo.

View on GitHub

Publications

10 publications · 52+ citations Live from Scholar · Google Scholar · ResearchGate

2024
Q1 · IF 11

Blockchain-Integrated Sustainable Supplier Selection and Order Allocation: A Hybrid BWM-MULTIMOORA and Bi-Objective Programming Approach

A. Vaezi, E. Rabbani, S.A. Yazdian

Journal of Cleaner Production, 444, 141216

52+ Citations Q1 · IF 11
Read Paper
2026

Integrating Fuzzy MCDM Priors into Deep Reinforcement Learning for Sustainable Supplier Selection and Dynamic Order Allocation

A. Vaezi, E. Rabbani, Prof. G. Bruno

International Journal of Production Economics (IF 7.9)

Under Review
View Code
2026

xAI-Driven Understanding and Mitigation of Hallucinations in MLLMs: A Cross-Domain Study on Medical and Industrial Visual Inputs

A. Vaezi

CLiC-it 2026 (Italian Conference on Computational Linguistics)

In Preparation
View Code
2026

A Stochastic User Equilibrium Model for Multimodal Transportation in Tehran

A. Vaezi, S. Alitanlo, A. Naghedi Ahmadi, C. Pronello

Nested-Logit Stochastic User Equilibrium · Multimodal Transport

Under Review
View Code
2025

SSRN & ResearchGate Working Papers

A. Vaezi et al.

6 additional preprints on SSRN & ResearchGate

  • Generalised Cost Analysis of Urban Mobility Options
  • Comprehensive Analysis of Shared E-Scooter Micromobility in Turin
  • Legal Challenges in LLM Deployment: GDPR & EU AI Act
  • Enhancing Leadership Effectiveness via ML (Random Forest & K-Means)
  • Urban Heat Risk Mapping in NYC Using GIS & Fuzzy AHP
  • Urban Building Energy Modelling (UBEM) of Turin's Statistical Area 7
  • Context-Aware North Star Metric Selection for Startups (BWM-TOPSIS)
View All on ResearchGate

Certifications

40+ certifications across ML, Data Science, RL, and professional development.

Stanford University

2023
Machine Learning Specialization
  • Supervised Machine Learning: Regression & Classification
  • Advanced Learning Algorithms
  • Unsupervised Learning, Recommenders, Reinforcement Learning

IBM

2023
Data Science Professional Certificate 9 courses

Google

2023
Data Analytics Professional Certificate 8 courses

Harvard University

2023
CS50x: Introduction to Computer Science

Udemy

2025
4× Advanced Reinforcement Learning Courses

IEEE

2025
Graduate Student Member #100632219

MathWorks

2023
Programming with MATLAB

Academic & Professional

2022–2026
I3P / PoliTo: Therness Hackathon LinkedIn: Git Workflows DataCamp: OOP in Python Great Learning: Data Analytics K.N. Toosi: SPSS Certificate Tehran Tech: Python Programming Urmia University: GAMS Amirkabir: Agile PM Isfahan: 20th IE Conference Sharif University: Certificate 10th WSS: Winter Seminar (SUT)

Sandbox

A few things I tinker with in my free time.

Edge AI

Deploying ML models on Raspberry Pi and Jetson Nano for real-time inference at the edge.

Discover more

RL Environments

Custom Gymnasium environments for supply chain simulation and multi-agent coordination.

Discover more

MuJoCo Sim

Physics-based locomotion simulations in MuJoCo with domain randomization for sim-to-real transfer.

Discover more

Data Viz

Interactive dashboards with Plotly, Streamlit, and Power BI for supply chain KPIs.

Discover more

System Dynamics

OpenCLD: my open-source Python library for system dynamics simulation, causal loop diagrams & stock-flow.

Discover more

Writing

Writing LaTeX papers, LinkedIn posts on IoT and AI, and documenting my research journey.

Discover more

Contact Me

I am always open to discussing research collaborations, thesis opportunities, and exciting projects. Feel free to reach out!

Messages are sent to my email. I typically reply within 24–48 hours.