Title: Jun Hur | Software Engineer · Reliable Cyber-Physical Systems
Open Graph Title: Jun Hur | Software Engineer · Reliable Cyber-Physical Systems
X Title: Jun Hur | Software Engineer · Reliable Cyber-Physical Systems
Description: Jun Hur — software engineer with 5+ years building production systems across robotics, fintech, and healthcare, focused on the reliability of cyber-physical systems: cross-layer fault diagnosis, failure prediction from operational telemetry, and simulation-based validation before deployment.
Open Graph Description: 5+ years building robot fleet control, fintech, and healthcare systems — researching how to make cyber-physical systems diagnosable across layers, predictable from telemetry, and validated in simulation before deployment.
X Description: 5+ years building robot fleet control, fintech, and healthcare systems — researching how to make cyber-physical systems diagnosable across layers, predictable from telemetry, and validated in simulation before deployment.
Mail addresses
hurjun96@gmail.com
Opengraph URL: https://hurjun.github.io/
Domain: hurjun.github.io
| author | Jun Hur |
| theme-color | #0a192f |
| og:type | website |
| og:image | https://hurjun.github.io/asset/face.jpeg |
| twitter:card | summary |
| twitter:image | https://hurjun.github.io/asset/face.jpeg |
Links:
| Jun Hur | https://hurjun.github.io#hero |
| Research | https://hurjun.github.io#research |
| About | https://hurjun.github.io#about |
| Skills | https://hurjun.github.io#skills |
| Experience | https://hurjun.github.io#experience |
| Projects | https://hurjun.github.io#projects |
| Education | https://hurjun.github.io#education |
| GitHub | https://github.com/hurjun |
| https://www.linkedin.com/in/jun-hur-a3b0b7159/ | |
| KCC paper | https://hurjun.github.io/asset/kcc2024-kiise-nursing-record-paper.pdf |
| Xpanner | https://xpanner.com/ |
| GoLe Robotics | https://golerobotics.com/home |
| Niflers | https://niflers.com/ |
| DOGU Robotics | https://dogu.xyz/ |
| golerobotics.com ↗ | https://golerobotics.com/home |
| crisk.co.kr ↗ | https://www.crisk.co.kr/ |
| enesg.co.kr ↗ | http://www.enesg.co.kr/ |
| github.com/hurjun | https://github.com/hurjun |
| MAPF Fleet — Multi-Robot Construction Site Simulator GitHub ↗ 3D multi-robot fleet simulation on multi-floor sites with a MAPF engine written from scratch in TypeScript (windowed cooperative A*, Conflict-Based Search). Zero collisions across 66 headless benchmark runs / 21,000+ ticks. TypeScriptNext.jsReact Three FiberMAPF | https://github.com/hurjun/mapf-fleet |
| PPE Watchman — Edge-to-Cloud Site Safety Monitoring GitHub ↗ Edge-to-cloud pipeline for site-safety alerts: YOLOv8 edge clients (stock weights, PPE model pluggable) send events to a FastAPI hub, which fans them out over WebSocket with backpressure. 33 tests; detection accuracy not evaluated. FastAPIYOLOv8Next.jsWebSocket | https://github.com/hurjun/mlops |
| Transformer from Scratch GitHub ↗ Attention Is All You Need re-implemented from primitives in PyTorch. Reaches 98.2% exact sequence match on a synthetic reversal task (CPU, ~90 s). In a separate 1,200-step ablation, removing positional encoding drops exact match from 91.2% to 0%. PyTorchTransformersNLP | https://github.com/hurjun/attention_is_all_you_need |
| xLSTM from Scratch GitHub ↗ sLSTM and mLSTM cells (Beck et al., NeurIPS 2024) implemented from scratch in PyTorch. On a 64-step recall task, xLSTM reaches 100% accuracy while a vanilla LSTM stays at chance (≈6.25%), even with up to 2× the parameters. PyTorchRNNSequence Modeling | https://github.com/hurjun/xLSTM |
| Speculative Decoding Lab GitHub ↗ Speculative decoding with distribution-preserving acceptance sampling, implemented from scratch. Up to 1.80× lossless speedup on Apple Silicon (DistilGPT-2 drafting, GPT-2-large verifying), with output token-identical to greedy decoding. PyTorchLLM InferenceGPT-2 | https://github.com/hurjun/speculative-decoding-lab |
| MediVoice — AI Speaker Nursing Documentation GitHub ↗ Voice-first nursing documentation: an AI speaker asks assessment questions (TTS), transcribes the answers (STT), and normalizes them into structured records. Implementation of the KCC 2024 paper (3rd of 4 authors): documentation time −70% on first use, −96% on repeated use, on simulated records. ReactTypeScriptWeb Speech API | https://github.com/hurjun/stt-nursing-system |
| Stock Forecast Benchmark GitHub ↗ Leakage-controlled benchmark of 9 forecasting models, from ARIMA to Transformer, on 1962–1992 → 1993–2000 stock data. Prophet ranks first (RMSE 160), with an analysis of why the deep-learning models underperform. PythonPyTorchForecasting | https://github.com/hurjun/stock-forecast-benchmark |
| KCC 2024 paper (PDF, in Korean; English summary on pp. 1–2) | https://hurjun.github.io/asset/kcc2024-kiise-nursing-record-paper.pdf |
| Code (stt-nursing-system) | https://github.com/hurjun/stt-nursing-system |
| github.com/hurjun | https://github.com/hurjun |
| linkedin.com/in/jun-hur-a3b0b7159 | https://www.linkedin.com/in/jun-hur-a3b0b7159/ |
Viewport: width=device-width, initial-scale=1.0