01 / ENTROPYPHYSICAL SIGNAL
01
Turn randomness into a secure flow.
Physical entropy, AI-hybrid TRNG and a layered security approach for verifiable process design.
AI-Driven Computational Electromagnetics · AI-CEM
Hasan Yiğit
I build the full computational stack of electromagnetic science — from Maxwell's equations to working systems: differentiable FDTD solvers, AI-driven inverse design, scientific software and the GPU/HPC infrastructure that makes electromagnetic design fast, reproducible and deployable.
EducationAI M.Sc. 3.93 · EEE B.Eng. 3.73
AnchorDifferentiable FDTD · AI inverse design · Scientific software
DirectionPhD candidate · AI-CEM research programme
00 / SIGNAL FLOW
Good systems make the invisible readable.
Each piece of work starts from the same question: how do we turn a complex physical or numerical signal into a reliable decision and a reusable system?
source model decision
↓ Scroll to explore the cards
↗
02 / FIELDFDTD · NUMERICAL SOLVE
02
Sample the field, see the behaviour.
Computational electromagnetics make high-frequency systems, parameters and wave behaviour numerically readable.
03 / VISIONAI · IMAGING
03
Bring images closer to a measurable decision.
Computer vision pipelines that turn raw visual data into auditable, meaningful outputs in health and retail.
04 / LEARNINGADAPTIVE SYSTEM
04
Place the model inside a real system.
Adaptive architectures that bring data, simulation and infrastructure into one workflow — from research to a working product.
01 / THE STACK — ONE RESEARCH DIRECTION
Computational Electromagnetic Systems.
One stack, four layers.
The backbone of the AI-CEM research programme: a four-layer stack running from Maxwell's equations to working systems on GPU/HPC. Each layer enables the one above it.
01 / PHYSICS
Computational Electromagnetics CEM
Numerical solution of Maxwell's equations with FDTD, FEM and MoM. The foundation for high-frequency systems, wave propagation and inverse problems.
02 / INTELLIGENCE
AI & Scientific Machine Learning AI · SciML
Differentiable programming, neural operators (FNO/GNO), curvature-aware optimization and graph-based search. The layer where AI is the method, and EM is still the problem.
03 / SOFTWARE
Scientific Software Software
Open-source solvers, differentiable frameworks, benchmark suites, datasets and reproducible pipelines. Every published method ships as a concrete software artifact.
04 / INFRASTRUCTURE
GPU / HPC Infrastructure Infrastructure
GPU and distributed computing, memory-aware design, MLOps and verifiable systems. The engine that lets solvers and models run at real scale.
05 / APPLICATION
Electromagnetic Systems EM Systems
Antennas, microwave components, 6G systems, metasurfaces, computational imaging and RF sensing. The concrete output of the whole stack.
// One research direction. Multiple engineering capabilities. — From Maxwell's equations to working systems.
02 / RELATED WORK
The adjacent fields around the main stack.
Inverse scattering, microwave imaging and RF tomography. Medical image processing and computer vision live here as a physics-grounded imaging bridge of AI-CEM.
Physical-layer security, RF fingerprinting, EM side-channel analysis and trusted AI for EM systems. Secure-infrastructure background can turn into a research thread here.
JAX/PyTorch-based EM solvers, benchmark suites and datasets. Concrete software artefacts that accompany every published method.
QuevaTech for physical entropy, key management and verifiable processes. The discipline of carrying research output into product and industrial context.
NEWS — 01
How Do We Know Tomorrow's Weather?A talk on numerical modelling, chaos and AI.
↗
03RESEARCH PROGRAMME
RESEARCH PROGRAMME
I think about a field not in isolation,
but through the whole stack.
The AI-CEM research programme is a four-layer stack running from Maxwell's equations to working systems. Physics, intelligence, software and infrastructure all serve the same scientific question: how can electromagnetic problems be solved faster, more accurately and more autonomously?
3.93AI M.Sc. GPA / 4.00
3.73EEE B.Eng. GPA / 4.00
2022—research assistant
SELECTED WORK
From idea to prototype.
01 / SECURE INFRA↗
QuevaTechQuevaTech
Physical entropy, AI-hybrid TRNG, key/password management and secure infrastructure ideas for public proof of process immutability.
02 / IMAGE PROCESSING↗
Image Processing Systems
Deep learning and computer vision across health, oral/radiological images, retail analytics and in-store observation scenarios.
03 / COMPUTATIONAL ENG.↗
Computational Electromagnetics
High-frequency system simulation, FDTD, parameter sweeps and numerical engineering for high-frequency design.
ACADEMIC CV
Research, product and infrastructure
on one page.
request
A scannable summary of academic history, publication evidence and topics open for collaboration. A PDF résumé can be prepared for a specific role or project.
Open CV ↗04 / CONTACT
If you have a question,
an idea or a reason
to think together.
For talks, academic collaboration or technology-focused projects, a short message is enough.