AI & Robotics Researcher · Tehran
Hamed Hosseini
PhD in AI & Robotics, University of Tehran. Postdoc at TAARLab — the Human–Robot Interaction Lab — working on robotic grasp detection and scene understanding; I also lead applied AI teams in finance.
About
I do research at the point where vision meets manipulation: teaching robots to understand objects as geometric primitives and infer stable grasp poses from RGB-D data. My doctoral work — primitive-shape scene abstraction for grasp detection — was implemented in simulation and validated on a real parallel delta robot with a custom gripper. I now continue this work as a postdoctoral researcher at TAARLab, the Human–Robot Interaction Laboratory at the University of Tehran.
Parallel to research, since 2018 I have led a data science team in banking: fraud detection, credit evaluation, and LLM-based customer systems, built on Spark and Kafka pipelines.
- BSc
- Electrical Engineering · Amirkabir University of Technology · 2015
- MSc
- Electrical Engineering, Control · Amirkabir University of Technology · 2017
- PhD
- Artificial Intelligence & Robotics, University of Tehran · 2017–2025
- Now
- Postdoc · TAARLab — Lecturer (AI in Robotics) · Data Science Team Lead
Research
Robotic grasping
Geometry-aware grasp pose inference for finger grippers, from RGB-D observation to robot action.
Scene abstraction
Decomposing objects and scenes into 3D primitives — the representation behind my grasp-detection work.
Embodied AI
Rearrangement and relocation tracking in simulated kitchen environments (AI2-THOR).
Computer vision
Detection, segmentation, RGB-D perception, and robust multimodal models.
Applied AI
LLM systems, RAG, fraud analytics, and financial machine learning in production.
Publications
Full list on Scholar ↗Journal
Dynamic Buffers: Cost-Efficient Planning for Tabletop Rearrangement with Stacking
Barghi, Hosseini, Ghasemi, Masouleh, Kalhor — IEEE Trans. Automation Science and Engineering · YouTube ↗
Embodied AI for Kitchen Scene Rearrangement: A Deep Spatio-Temporal Approach
Nasr Esfahani, Hosseini, Masouleh, Kalhor — Intelligent Service Robotics · YouTube ↗
Multi-Modal Robust Geometry Primitive Shape Scene Abstraction for Grasp Detection
Hosseini, Koosheshi, Masouleh, Kalhor — IEEE Access · DOI · video
Conference
Push-Placement: A Hybrid Approach Integrating Prehensile and Non-Prehensile Manipulation for Object Rearrangement
Sadeghinejad, Barghi, Hosseini, Masouleh, Kalhor — accepted · arXiv ↗
Architecture Design and Branching Strategies for Sound Event Detection Using Cross Separation Index
Vaghef Davari, Hosseini, Beigy, Masouleh, Kalhor — 34th ICEE
Learning Post-approach Robotic Grasping using Convolutional Neural Networks
Koosheshi, Hosseini, Masouleh, Haeri Yazdi, Kalhor — 13th RSI ICRoM · oral
Scene Understanding in Pick-and-Place Tasks: Analyzing Transformations Between Initial and Final Scenes
Ghasemi, Hosseini, Koosheshi, Masouleh, Kalhor — 32nd ICEE · YouTube ↗
AI-Driven Relocation Tracking in Dynamic Kitchen Environments
Nasr Esfahani, Hosseini, Masouleh, Kalhor, Sajedi — 14th ICCKE · YouTube ↗
AGILE: Approach-Based Grasp Inference Learned from Element Decomposition
Koosheshi, Hosseini, Masouleh, Haeri Yazdi, Kalhor — 11th RSI ICRoM · YouTube ↗
Improving the Successful Robotic Grasp Detection Using Convolutional Neural Networks
Hosseini, Masouleh, Kalhor — ICSPIS
PARSIAN 2017 Extended Team Description Paper
Rahimi, Shirazi, Arfaee, Gholian, Zamani, Hosseini, et al. — RoboCup
Experience
Postdoctoral Researcher · TAARLab
Human–Robot Interaction Laboratory, University of Tehran — robotic grasping, scene abstraction, and embodied AI.
Lecturer · AI in Robotics and Mechatronics
University of Tehran. Course materials ↗ · YouTube ↗
Data Science Team Lead
Pouya Co. · Banking solutions — fraud detection, credit evaluation, LLM support systems, Spark/Kafka pipelines.
Research Assistant · TAARLab
Human–Robot Interaction Laboratory, University of Tehran — grasping, scene understanding, deep learning.
Research Assistant · Robotics & Signal Labs
Amirkabir University of Technology — small-size robot soccer (RoboCup 4th place 2017), control, biosignal ML.
Workshops & Talks
Details on LinkedIn ↗New Approaches in Robotic Grasping and Object Manipulation
Workshop at the 28th Elecomp International Exhibition, Tehran — modern grasp detection and manipulation methods. With Dr. Mehdi Tale Masouleh and Hamed Ghasemi.
HRI Lab Projects at the 2nd AI & IoT Conference (INOTEX)
Presented the Human–Robot Interaction Laboratory's robotic-grasping projects and results at Pardis Science & Technology Park, with fellow lab researchers.
Panel — Motion Control Technologies in Robotics: From Machine Learning to Industry 4.0
IranOpen robotics exhibition — presented HRI Lab research on robot motion control, autonomous navigation, physical interaction, and robotic grasping, with several projects now in industrial use. With Dr. Mehdi Tale Masouleh, Dr. Ahmad Kalhor, Dr. Vahid Bahrami, and Dr. Hamed Ghasemi.
Robotic Grasping Through Time: Progress, Challenges, and New Frontiers
Workshop at ICRoM 2024, Amirkabir University of Technology — from classical grasp detection to AI-driven methods: 3-DOF Delta robots, reinforcement learning, scene understanding, graph neural networks, and VLA models (π0, OpenVLA). With Dr. Mehdi Tale Masouleh (chair) and Hamed Ghasemi.
New Approaches in Robotic Grasping: From Deep Learning, Reinforcement Learning to Graph Neural Networks
Online workshop at 14th ICCKE — six years of HRI Lab research: geometric and intelligent grasp detection, RL for data insufficiency, graph modeling, simulators, and a custom gripper on a 3-DOF Delta robot, plus Transformers in manipulation (GPT-4V, RT-1/RT-2). With Dr. Mehdi Tale Masouleh, Hamed Ghasemi, and Hamed Hosseini.
AI in Medical Sciences
Seminar at the ENT Research Center, Amir Alam Hospital, Tehran — how AI transforms diagnostics, treatment planning, and patient care, including ChatGPT and Gemini in research. Speaker alongside Prof. Hamid Berenji (NASA researcher).
Deep Learning Perspectives on Robotic Grasping
Workshop at AISP 2024 — deep-learning-, reinforcement-learning-, and graph-based grasp detection, and Transformer applications in robotic manipulation. With Dr. Mehdi Tale Masouleh and Hamed Ghasemi.
TEDx Talk
Invited TEDx talk — AI and robotics: my research story. Paper videos and media highlights on LinkedIn ↗
Robotic Grasping: A Deep Learning Approach
Workshop at 11th RSI ICRoM — six years of gripper-design and grasp-detection experience: geometric, primitive-shape, graph-based, RL, point-cloud, and imitation-learning methods on a 3-DOF Delta robot with a custom two-fingered gripper, plus Transformers in manipulation (GPT-4V, RT-1/RT-2). With Dr. Mehdi Tale Masouleh, Hamed Hosseini, and Hamed Ghasemi.
