2025
KTH Royal Institute of Technology
Thesis on efficient and generalizable machine learning for seizure detection from EEG.
An AI and machine learning engineer working with time series, sensor data and LLM systems.
Get in touch →I’m Zheyun, an AI and machine learning engineer based in Stockholm. I build models for time series and sensor data, and LLM systems that turn messy input into something a product can use.
I’ve worked on seizure detection from EEG, haptics generated from plain language, fault forecasting for industrial pumps and robot control with reinforcement learning. I like the whole path, from the raw signal to the code around the model.
Hover the underlined words
Hover a card
Four places, one thread. Noisy signals in, something reliable out.
From my MSc thesis at KTH. Time-series classifiers that stay light and keep their accuracy when they move between datasets.
ROCKET, catch22, signal processing, multi-channel sensor data
PyTorch, scikit-learn, NumPy, pandas, representation learning, deep generative models
LangGraph agents, MCP, RAG with hybrid retrieval, agent harnesses with tool calling, guardrails and graders, multi-model routing
Git, CI/CD, Docker, GCP, Azure
Python, SQL, C++
Chinese (native), English (fluent), Swedish (conversational)
Thesis on efficient and generalizable machine learning for seizure detection from EEG.
A semester abroad during my bachelor’s.
Thesis on a computer vision system for PCB defect detection.