M.Sc. student in Wireless Communications Engineering at the University of Oulu. Previously M.Sc. Micro/Nano Electronics (Tarbiat Modares) and B.Sc. Electrical Engineering (Tabriz). Two years of industry R&D at Sepidgaman Parseh in Tehran.
I’ve always worked in electronics, but not in the narrow sense of just designing circuits. For me, it’s one continuous system that starts from physics and extends all the way to working machines — from the physics of devices, to circuit and hardware design, to embedded systems and higher-level algorithms.
What interests me is the connection between layers. The same electron that follows quantum mechanics in a material ends up shaping a signal on a PCB, a control loop in firmware, or data going into a model. I haven’t really switched fields over time — I’ve just worked at different levels of the same system, trying to understand how these layers fit together and influence one another.
My work reflects that. I’ve built optical and electromagnetic simulations, designed circuits and PCBs, developed embedded systems, and worked on perception and machine learning problems. Many of these ideas have been tested in real systems, from robotics platforms to sensing and communication setups, where constraints and imperfections matter as much as theory. Different tools, same mindset: understand the system from the ground up, and make sure it actually works in practice.
I care about physical correctness. Whether it’s checking energy conservation in an optical model or validating a hardware system on the bench, I don’t like treating simulations or abstractions as black boxes.
This portfolio is organized the same way I think — as a stack. From physics and simulation at the bottom, through hardware and embedded systems, up to higher-level intelligence. Each layer builds on the previous one, and understanding their interaction is a central part of how I work.
Two areas I am actively reading and thinking about.
Diffractive Deep Neural Networks (D2NN, Lin et al. 2018) perform inference entirely in the optical domain: phase masks replace weight matrices, and computation happens at the speed of light through free-space diffraction. My interest is in the physical feasibility gap — the distance between a network that converges in simulation and one that can actually be fabricated. This involves sampling theory (ASM band-limit), energy conservation, photon-budget analysis, SNR under detector noise, and Monte Carlo fabrication tolerance analysis. More broadly, I am interested in what it means to implement neural network operations with physical substrates rather than transistors.
The Xu et al. 2024 ISAQN framework (Integrated Sensing And Quantum communication Networks) demonstrates that quantum key distribution (CV-QKD) and distributed optical fiber sensing (DOFS) can coexist on the same fiber through frequency-division multiplexing. The sensing channel uses Rayleigh backscattering for vibration sensing, with pilot-tone phase recovery bringing it to the quantum noise limit — meaning the sensing precision is bounded by quantum mechanics, not classical electronics. I am interested in this convergence: using the same infrastructure for both secure communication and physical-world sensing, and the signal processing that makes both work simultaneously without mutual degradation.
Sixteen projects across four levels of the electronics stack.
Device-level physics, electromagnetic simulation, quantum mechanics, photonic design — the foundation.
ASM propagation engine with five quantitative checks: band-limit, energy conservation, photon budget, SNR, and Monte Carlo fabrication tolerance.
Analytical SOI mode demultiplexer — EMT, EIM, MMI self-imaging, full 2D field reconstruction across 300 nm bandwidth. Validated against Lumerical FDTD; scripts available upon request.
Bound-state-in-the-continuum Si nanocube resonator (Q > 1000) for label-free cancer vs. non-cancer cell detection. COMSOL FDTD + Kramers–Kronig conversion + SVM classification.
Semiclassical Maxwell-Bloch FDTD for gain-modulated MSM waveguides and microcavities. Gold modelled with Lorentz-Drude; ~60% resonance amplitude increase demonstrated.
Faithful reproduction of Slavcheva et al. semiclassical FDTD for VCSEL lasing at 850 nm and 1.29 μm. 1D Yee grid, Mur ABC, gain saturation.
COMSOL Multiphysics 2D PhC simulation of all-optical NOT, AND, OR gates at λ = 1.55 μm based on Rani et al. 2015. Interference-based switching without nonlinearity.
Luttinger-Kohn Hamiltonian solver for GaAs/Al₀.₅Ga₀.₅As double quantum well (25 Å / 15 Å barrier). HH/LH Gaussian wavepacket dynamics in MATLAB.
Analog and mixed-signal design, PCB, RF/inductive systems — turning physics into hardware.
~1 MHz magnetically-coupled near-field link, push-pull LC driver, ASK/OOK modulation, beaconed TDMA MAC for up to 8 nodes over 1–3 m. University of Oulu FabLab.
ATmega16-based impedance meter using AC bridge at 10 kHz. Peak detector and zero-crossing detector measure magnitude and phase; polar-to-rectangular conversion in firmware.
Full SMPS design cycle: topology selection, magnetics, control loop compensation (Bode analysis), PCB layout, and bench validation. Final-year undergraduate project.
RTOS design, microcontroller firmware, communication protocols, edge computing.
Contract-grade 6-task RTOS datalogger. 100 Hz IMU, 5 Hz GPS, RS232 telemetry, raw SD sector writes with triple-redundant pointers. Paid client project.
HTML/JS frontend + Python backend modeling proton-exchange and thermal-oxidation processes. Accepts Excel inputs or live sensor readings; outputs process curves in real time.
Face recognition + 9-class MediaPipe gesture recognition, dual-threaded Python client, TCP to ESP32 server. Real-time identity and gesture dispatch over Wi-Fi.
Perception, machine learning, autonomous systems — intelligence built on all lower layers.
64-ch EEG at 250 Hz. Three classifiers: 1D CNN (83.8%), EEGNet (83.4%), CSP+SVM (81.2%). BCI Competition 2005 Dataset I with 5-fold stratified cross-validation.
12 Hz chest-worn accelerometer. Two-stage hierarchical classifier (Stage-1 coarse: 4 groups; Stage-2 fine: 12+ activities). Handcrafted time/frequency features, Random Forest · XGBoost · SVM ensemble for clinical activity monitoring.
Raspberry Pi 4 + PiCamera2. Three sign-detection backends (template matching → custom CNN → YOLOv5), lane following via binary threshold + centroid, UDP JPEG stream.
Built from scratch at age 15. Dead-reckoning odometry, 8-sensor ring, 10×10 occupancy grid, Dijkstra planning, boustrophedon coverage. Arduino Mega 2560.
Operated independently on a project-by-project basis after leaving full-time employment, commanding higher rates than the preceding salaried role. Delivered the STM32H FreeRTOS datalogger system (5 field-iterated versions) under a paid contract, taking sole ownership of firmware architecture, hardware bring-up, and field debugging. Took on additional embedded and hardware design engagements where technical depth and autonomy were the primary requirements.
Worked in an ISO Class 7 cleanroom environment on microbolometer FPA and ROIC inspection systems for thermal imaging. Designed and debugged analog front-ends and mixed-signal PCBs for biomedical instrumentation using Altium Designer. Developed firmware for STM32, AVR, and PIC microcontrollers under production constraints. Wrote Python and MATLAB automation tools for measurement data acquisition and post-processing. Worked closely with Cadence OrCAD/PSpice for circuit simulation. Engagement was primarily through project-based contracts across both the Tehran and Shiraz offices.
A different kind of engineering — not for a client, not for a grade.
Chista is a serious competitive robotics team based in Shiraz. I joined in 2014 and have been the hardware and embedded programmer since — responsible for microcontroller firmware, sensor integration, motor drivers, and real-time control systems on competition robots. The team competes at a high level in regional and international leagues, but none of us are paid. It is shared passion: the kind of work where you stay up until 3 a.m. before a competition not because you have to, but because you cannot leave a problem unsolved.
Two projects in this portfolio grew directly out of Chista work: the Autonomous Differential-Drive Robot — built from scratch when I was around 15 — and the Autonomous Car Vision System, which applied the same embedded-plus-perception approach to a track-following vehicle. The competition history below spans a decade of that work.
Certificates of participation and awards (where applicable) are available on request.
| Competition | Year |
|---|---|
| Autonomous Vehicles Tehran | 2025 |
| Asia Pacific Robot Contest | 2018 |
| Iran Open Robocup | 2016, 2017 |
| AUT Cup (Amirkabir University) | 2017 |
| FIRA Cup | multiple |
| Shiraz Open | 2015 |
Additional regional competitions not listed. Documentation available on request.