Computer Vision & Edge AI · From training to inference, running models on real hardware
I work on the engineering side of computer vision — model development, inference optimization, and edge deployment. I believe a model's value is measured by how it performs on real hardware.
Vision Models Object detection, segmentation, OCR, face, ReID, fine-grained classification. YOLO family, CRNN/CTC, multimodal models — from paper to production.
Model Engineering ONNX, TensorRT, Triton. Pruning, distillation, quantization. CV workflow toolkits spanning data processing, model export, inference, and evaluation.
Edge Deployment NVIDIA Jetson, Huawei Ascend, SOPHON, Horizon, Rockchip RKNN. C++ multi-stream inference, model acceleration, cross-platform adaptation.
LLM & Multimodal LLaMA-Factory LoRA fine-tuning, instruction-data & prompt engineering. vLLM serving — LoRA direct or merged weights. Evaluation & benchmarking via OpenAI-compatible APIs.
| Project | Description |
|---|---|
| DataFlow-CV | 🌊 Everything your model doesn't do. Analyse, convert, visualize, evaluate — a single CLI for all CV data. |
| InferForge | 🔨 From kernel to service — InferForge forges any model (CV → LLM → Agent) into production |
| llamafactory-lora-vllm | 🦙 LoRA fine-tuning with LLaMA-Factory → high-performance inference deployment via vLLM |
| ZJCV · Organization | 📚 DL paper reproductions & training frameworks — classification, video action recognition, distillation & pruning (archived) |
| ZJDoc · Organization | 📖 Engineering guides & docs — Git, Linux, deployment, vision & tooling (archived) |