cv

Generated from my Overleaf LaTeX CV, so this page and that document never disagree.

Contact Information

Name Mohammad Taha Fakharian
Professional Title Ph.D. Student
Email mohammad.fakharian@oist.jp
Location 1919-1 Tancha, Onna-son, Okinawa 904-0495

Professional Summary

Ph.D. student at OIST working at the intersection of computational neuroscience and machine learning, on how learning systems build world models from sensory data and what makes those models generalize.

Experience

  • 2026 - 2026
    Ph.D. Research Rotation — [Prof. Jun Tani](https://www.oist.jp/research/research-units/cnru/jun-tani)
    [Cognitive Neurorobotics Research Unit](https://www.oist.jp/research/research-units/cnru), OIST
    Developed a reproducible evolutionary-developmental extension of Curiosity-driven development of action and language in robots through self-exploration (Tinker, Doya, and Tani, 2026), evolving a compact meta-controller to regulate exploration and lifetime learning.
  • 2026 - 2026
    Ph.D. Research Rotation — [Prof. Tomoki Fukai](https://www.oist.jp/research/research-units/ncbc/tomoki-fukai)
    [Neural Coding and Brain Computing Unit](https://www.oist.jp/research/research-units/ncbc), OIST
    Investigated how short-term synaptic plasticity and recurrent connectivity support dynamic persistent activity and temporal associative-memory sequences, extending Theory of Coupled Neuronal–Synaptic Dynamics (Clark and Abbott, 2024) and Extended Temporal Association Memory by Modulations of Inhibitory Circuits (Haga and Fukai, 2019) in rate and spiking-network simulations.
  • 2025 - 2025
    Ph.D. Research Rotation — [Prof. Kenji Doya](https://www.oist.jp/research/research-units/ncu/kenji-doya)
    [Neural Computation Unit](https://www.oist.jp/research/research-units/ncu), OIST
    Developed a spiking basal-ganglia model of dopamine-modulated plasticity to study synaptic credit assignment and circuit-level temporal-difference learning, building on A Hardwired Neural Circuit for Temporal-Difference Learning (Campbell et al., 2025); presented this work at the Winter Workshop 2026 on Mechanism of Brain and Mind.
  • 2025 - 2025
    Remote Research Collaborator — [Prof. Geoffrey Goodhill](https://neuroscience.wustl.edu/people/geoffrey-goodhill-phd/) and Daniel Zavitz
    [Goodhill Lab](https://www.goodhill.org/), Washington University in St. Louis
    Investigated how excitatory–inhibitory imbalance in recurrent-network models can inform computational accounts of psychiatric disorders. Admitted to Washington University’s Systems Science and Mathematics Ph.D. program to join the lab; enrollment was precluded by the June 2025 U.S. visa suspension.
  • 2024 - 2025
    Research Assistant — [Prof. Shervin Safavi](https://www.digs-bb.de/research/research-groups/shervin-safavi)
    [Computational Machinery of Cognition Lab, TU Dresden](https://shervinsafavi.github.io/cmclab/)
    Investigated how recurrent neural network architectures shape internal dynamics across cognitive tasks from the NeuroGym library. Trained recurrent models with different architectures and compared their state-space dynamics using empirical dynamic modeling (EDM), methods based on the generalized Takens theorem, and dynamical similarity analysis (DSA).
  • 2023 - 2024
    B.Sc. Thesis Researcher — [Prof. Mohammadreza Abolghasemi](https://profile.ut.ac.ir/en/~dehaqani); co-supervised by [Prof. Timothée Masquelier](https://cerco.cnrs.fr/pagesp/tim/index.lab.htm)
    [Convergent Technologies Research Institute, University of Tehran](https://nbic.ut.ac.ir/) / [CNRS](https://cerco.cnrs.fr/)
    Investigated delay learning in convolutional spiking neural networks using surrogate-gradient learning. Implemented learnable axonal, dendritic, and synaptic delays using dilated convolution with learnable spacings (DCLS) in ResNet-18-based spiking architectures for event-based datasets. Implemented selected components using Numba CUDA to gain hands-on experience with GPU programming and custom CUDA kernels.
  • 2023 - 2024

    Tehran, Iran

    Data Scientist
    [Tapsi](https://tapsi.ir/en)
    Helped establish a new data science team, applying deep-learning and generative-AI methods to improve customer experience on an online ride-hailing platform. Designed and implemented scalable recommender-system solutions for Tapsi’s products.
  • 2022 - 2022
    Research Assistant — [Prof. Azadeh Shakery](https://profile.ut.ac.ir/en/~shakery)
    [School of Electrical and Computer Engineering, University of Tehran](https://ece.ut.ac.ir/en/ece)
    Investigated graph-based approaches to hate-speech detection, with the goal of explicitly exploiting structural relationships between tokens rather than relying solely on sequence modeling with transformer architectures. Applied graph convolutional neural networks to text representations, and explored hybrid approaches combining graph-based modeling with BERT.

Education

Publications

Talks

Awards

  • 2025
    [CCN 2025 Travel Award ($500)](https://drive.google.com/file/d/1Ss78JnpSzyoujFWKYeA7dixdmJAfR1Yo/view?usp=sharing)
    [Cognitive Computational Neuroscience Conference](https://2025.ccneuro.org/)
  • 2024
    [Best data analysis project](https://drive.google.com/file/d/1by8q7j_CtWgrDgnzkCsSsvYn5PsepWVu/view?usp=sharing)
    [BR41N.IO Hackathon](https://www.br41n.io/Spring-School-2024)
  • 2024
    Ranked 1st among bachelor students of the Computer Engineering
    University of Tehran
  • 2019
    Ranked 96 (Top 0.1%) in National University Entrance Exam
    National Organization of Educational Testing (NOET)
  • 2019
    Received scholarship
    Supporter Foundation of the University of Tehran

Service

Teaching

Projects

  • [Dynamical RNNs](https://github.com/tahafkh/dynamical-rnns) — Cognitive Neurorobotics Project

    Built a from-scratch, Numba-accelerated Elman RNN with backpropagation through time, sequence-specific learnable initial states, and SGD/Adam optimization. Analyzed the resulting hidden-state dynamics and attractor structure.

  • [Crystalline](https://github.com/Crystaline-Coin/crystaline) — Decentralized File-Storage Cryptocurrency

    Developed a Python proof of concept for a peer-to-peer, incentive-backed file-storage network. Implemented the blockchain, transaction and signature validation, node networking, fee calculation, tests, and a hybrid Proof-of-Work/redefined Proof-of-Activity consensus mechanism.

  • [Oak](https://github.com/noushkia/Oak) — Marketplace Platform Backend

    Built and Dockerized a Java/Spring backend for an Amazon-style marketplace. Developed REST APIs for users, providers, commodities, ratings, comments, search, sorting, pagination, and recommendations; implemented JWT-protected authentication and GitHub OAuth sign-in.

  • [Smart Pot](https://github.com/noushkia/Smart-Pot) — Automated Plant-Watering System

    Built a three-controller Arduino/C++ system: a sensor board measures temperature and soil moisture, a master board applies the irrigation policy and displays measurements, and an actuator board drives the water pump by PWM. The controllers exchange data over Bluetooth serial communication.

  • [xv6 Kernel](https://gitlab.com/btk-os) — Extended xv6 Teaching Kernel (Five OS-Lab Projects)

    Progressively extended a fork of MIT’s xv6 with shell line-editing and shortcuts; custom process and filesystem system calls; a multi-level scheduler (round-robin, LCFS, and modified HRRN) with aging; semaphore synchronization; and mmap with demand paging, page-fault handling, and fork support.

  • [Socket Server](https://github.com/tahafkh/Socket-Server) — TCP Socket Servers Suite

    Implemented FTP, HTTP, and multi-user chat servers in C++ using low-level TCP sockets. Built FTP command/data channels with authentication and file upload/download, HTTP request parsing and routing for static/binary content, and chat-session and messaging workflows.

Certificates

  • [NeuroAI](https://portal.neuromatchacademy.org/certificate/d5ba21cb-ec0f-4183-b152-f4f9090e6ad0) - Neuromatch Academy (2024)
  • [Game Theory](https://www.coursera.org/account/accomplishments/certificate/FARJCC5ZBR5Z) - Standford University, The University of British Columbia (Coursera) (2023)
  • [Machine Learning Engineering for Production (MLOps) Specialization](https://www.coursera.org/account/accomplishments/specialization/VVKWAVGHNLDB) - DeepLearning.AI (Coursera) (2022)
  • [Natural Language Processing Specialization](https://www.coursera.org/account/accomplishments/specialization/3WLJG87GNUT3) - DeepLearning.AI (Coursera) (2022)
  • [Generative Adversarial Networks (GANs) Specialization](https://www.coursera.org/account/accomplishments/specialization/GDHUZ5XYGFVH) - DeepLearning.AI (Coursera) (2022)
  • [Deep Learning Specialization](https://www.coursera.org/account/accomplishments/specialization/TFDB7G4A94TX) - DeepLearning.AI (Coursera) (2022)
  • [Machine Learning](https://www.coursera.org/account/accomplishments/verify/EX8LXWX5UECM) - Stanford University (Coursera) (2022)
  • [Reinforcement Learning Specialization](https://www.coursera.org/account/accomplishments/specialization/JZSEY7WFLPK9) - University of Alberta (Coursera) (2022)

Skills

Programming: Advanced: Python, C++ Working knowledge: C, CUDA C++, Java, JavaScript, Bash, LaTeX Familiar: Lua
Machine Learning & Scientific Computing: PyTorch, JAX, NumPy, pandas, scikit-learn, TensorFlow, Keras, PySpark
Research & Development Tools: Git, Docker, Jupyter, Linux, Maven, Django, Spring
Agentic Workflows: LLM-assisted research and software development; tool-using agents, codebase analysis, and experiment automation

Languages

Persian : Native
English : Professional working proficiency **Academic IELTS: 7.5/9 [R:8.5, L:8.5, S:6.5, W:6.5] (Oct. 2023)**
Japanese : Beginner — currently studying _Genki I_ at OIST

Interests

Research: NeuroAI, Computational Neuroscience, Continual Learning, World Models, Dynamical Systems