Tehran تهران
IRThe beginning of a chapter.
A part of who I am.
CS AT SFU / MACHINE LEARNING INTERN
Mehrdad.
Learning to see.
Learning to build.
A curious mind somewhere between
humans, machines, and what comes next.
01 / SELECTED WORK
the things keeping me curious
Ideas into
something real.
Learning by building, asking better questions,
and getting a little further each time.
01CURRENT / MACHINE LEARNING INTERNSHIP
Huawei
Machine Learning InternMay 2025–Present · Markham, OntarioTraining robots to perceive, understand, and act with
reinforcement learning, VLMs, and VLAs.
CURRENTTHE LAB
02EXPERIENCE / SCIENTIFIC MACHINE LEARNING
Learning from the
Second author · Sole undergraduate on a six-person teamA multiphysics training framework for neural operators: up to
60% lower prediction error with 50% less training data.
2026READ PAPER
04PROJECT / GAME DEVELOPMENT
Huawei
Noah’s Ark Lab.
Machine Learning InternMay 2025–Present · Markham, OntarioTraining robots to perceive, understand, and act with
reinforcement learning, VLMs, and VLAs.
CURRENTTHE LAB
02EXPERIENCE / SCIENTIFIC MACHINE LEARNING
DeLTA Lab.
Machine Learning Research InternJan. 2025 – Jan. 2026 · Burnaby, BCMultiphysics learning and reproducible research: 500+ experiments supported by deterministic training checks, Dockerized GCP workloads, and shared artifact tracking. 2025–26THE LAB 03ICLR 2026 / RESEARCH AT DELTA LABLearning from the
physics of things.
Second author · Sole undergraduate on a six-person teamA multiphysics training framework for neural operators: up to
60% lower prediction error with 50% less training data.
2026READ PAPER
04PROJECT / GAME DEVELOPMENT
Galactic Blitz.
A different kind of space exploration. Take a look at the game in action. GAME DEVWATCH DEMO 05PROJECT / LANGUAGE MODELSOpti_LLM.
GPT-2, built from scratchA 124M-parameter GPT-2 implementation in PyTorch, with mixed precision and distributed-training optimizations that make compute go further. PYTORCHVIEW CODE 06PROJECT / COMPUTER VISIONRageVision.
A closer look at streaming videoA MobileNetV2 classifier trained to distinguish rage-labeled and non-rage frames in Twitch content through transfer learning. VISIONVIEW CODE 07PROJECT / NEUROEVOLUTIONNeuroDriver.
Learning to take the next turnA 2D driving simulation with neural networks built from scratch, evolving better steering and acceleration through selection, crossover, and mutation. PYTHONVIEW CODE02 / AN ENDURING CURIOSITY
I like knowing
what’s inside.
Old computers. New possibilities.
There’s something about
a machine you can take apart, understand, and put back
together.
a soft spot for old hardware.
01 / THE OUTSIDE
A familiar kind of hello.
A little beige box. A blinking cursor.
A whole world
waiting on the other side.
ASSEMBLEDEXPLODED
SCROLL TO OPEN IT UP EXTERIOR / 01LOOK CLOSER. THERE’S ALWAYS MORE.
OFF THE CLOCK / STILL CURIOUS
a few of my favourite rabbit holes
Not everything
needs a reason.
Some things are just good.
An open road. A good track. The
right key feel.
Form, function, and the open road.
Cars bring together so many things I like: design, engineering, and the feeling of going somewhere.
sometimes the long way is the right way.A SMALL CORNER OF MY WORLD
A place for everything.
The tools, the tiny cars, and the familiar things I come back to. A little sketch of my everyday setup.
FIG. 03 / THE EVERYDAY DESKIN ITS PLACE
EVERYDAYIN DETAIL
THE EVERYDAY COLLECTION
Good tools. Little distractions.
A keyboard I like typing on, a little room to think, and two cars that never leave the desk. Drag the slider or pick something above.
03 / THE HUMAN PART
a few places that stay with meA little bit
of everywhere.
I’m Mehrdad M. Zadeh, a computer science student at SFU, currently working as a Machine Learning Intern at Huawei’s Noah’s Ark Lab in Ontario.
My story has chapters in Tehran, Vancouver, and now Ontario. Different places, the same curiosity.
Vancouver ونکوور
BCBeautiful British Columbia
Still my favourite place.
Toronto تورنتو
ONA new setting.
Machine learning in nearby Markham.