ERC research · ongoing
HiWi at LMU Munich: T-MIGRANTS data-extraction tool
I am building a Python data-extraction and integration tool for the ERC-funded T-MIGRANTS project at LMU Munich (Department of Theatre Studies).
Yaoqiao Sha
M.Sc. Computer Science · Autonomous Systems · University of Stuttgart
Applied AI, knowledge graphs, edge-AI deployment. Currently in a cross-disciplinary M.Sc. lab applying knowledge graphs to architectural building representation. Open-source contributor to NumPy. Seeking a Master Thesis (industry-affiliated preferred) starting October or November 2026.
What I am working on now
- Lab Course AI: Knowledge Representation for Buildings (M.Sc., University of Stuttgart, 2026 in progress). Cross-disciplinary group project with architecture students. I proposed our group's pipeline architecture for an accessibility-checking KG + LLM system over BIM models: IFC → RDF (via IFCtoLBD) → SHACL rules → SPARQL queries → local LLM explanations. Implementation May to July 2026.
- Practical Course in Information Visualization: designing and shipping interactive data dashboards in D3.js.
- Continuing toward German B2 from a certified B1 (telc B1).
Selected work
Open source · scientific Python
NumPy UP031 refactor (tracking issue + multiple PRs)
I authored
numpy/numpy#30785,
the tracking issue coordinating NumPy's project-wide migration off legacy %-formatting
toward Python f-strings (under pyupgrade's UP031 rule). I contributed across
five pull requests spanning numpy/lib, numpy/_core,
numpy/_build_utils, numpy/f2py, and a final consolidation PR
(#31137, merged) that removed the
global UP031 exemption from ruff.toml.
Bachelor thesis · applied edge AI
YOLOv5 deployment on Huawei Ascend Atlas 200I DK A2
I deployed YOLOv5 inference end-to-end on the Huawei Ascend Atlas 200I DK A2 via the CANN / ATC toolchain: model conversion (PyTorch → ONNX → OM), on-device AIPP preprocessing and NMS placement, and a 200-trial benchmark protocol. mAP@0.5 = 43.0 on COCO 2017 val held with no accuracy loss vs. baseline. The Atlas's optimisation passes (operator fusion, selective INT8 quantisation) combined with the on-device pre and post-processing layout delivered ~4× throughput, ~76% lower latency (59.82 ms / image, 66.86 imgs/s) and ~15× higher energy efficiency (EER 97.07 vs. 6.31) versus a CPU baseline.
Research-assistant work · cross-functional
Quadruped robot inspection pipeline at oil & gas field stations
As a Student Research Assistant at the Computer Vision Lab, I worked cross-functionally with energy-domain engineers to integrate computer-vision modules into a quadruped robot's inspection-planning pipeline, applied for on-site monitoring at oil & gas field stations. Built simulation and control software in Unity / C# and 3ds Max; implemented routing & coverage algorithms in Java with a Test-Driven Development workflow.
Public repositories
SensorViewer
Android sensor reader (Kotlin). Mobile Computing M.Sc. coursework: sensors, BLE, beacons.
KotlinBLE_scanner
Bluetooth-LE device scanner (Kotlin). Mobile Computing coursework.
Kotlindiffusion2d
Python packaging exercise from Simulation Software Engineering coursework.
PythonRedisChannelForwarding
Forwarding script between Redis channels. Infrastructure tooling.
JavaScriptEducation
M.Sc. Computer Science, University of Stuttgart, Germany.
Specialisation: Autonomous Systems.
Oct 2025 to present (expected 2027)
B.Eng. Computer Science, China University of Petroleum (Beijing).
Sep 2020 to Jun 2024
Skills
Python (PyTorch, ONNX, OpenCV, NumPy, pandas, Jupyter) · Java · C++ · C# · JavaScript · Kotlin · Unity (C#) · 3ds Max · MySQL / MariaDB · Docker (M.Sc. coursework) · Git · Test-Driven Development · agile collaboration
Domain focus: Knowledge Graphs · NER / Entity Resolution · Generative AI & RAG (foundations) · Computer Vision · Edge / Applied AI · Autonomous Systems · Cross-disciplinary integration
Languages
Chinese: native · English: C1 (IELTS 7.0) · German: B1 (telc B1 certified, continuing toward B2)
What I am looking for
A Master Thesis (industry-affiliated preferred) starting October or November 2026, in applied AI, knowledge graphs, RAG, NER / entity resolution, or edge-AI deployment. Open to a company-hosted thesis with academic supervision from Stuttgart, or a research-institute arrangement with strong topic alignment. Long-term goal: full-time conversion after the M.Sc. via Blue Card.