onxx

OpenAI Whisper-compatible ASR server using NVIDIA Parakeet TDT 0.6B (ONNX). Super fast CPU/GPU transcriptions

Real-time, lightweight computer vision SDK for detecting eyeglasses using facial landmarks and machine learning. Built for the browser, Node.js, and edge devices via ONNX Runtime and MediaPipe FaceMesh.

GPU Pipeline for Model Import + Inference + Benchmark + Dashboard + Export

Agent-v1: Advanced LLM Agent with Multi-Layer Memory

Customer sentiment classification engine powered by INT8 quantized RoBERTa with calibrated negative recall (>90%) and real-time Kolmogorov-Smirnov drift monitoring.

Anime AI：支持截图识番、角色识别、番剧推荐、AI 对话与 RAG 知识库的助手平台。

A 711 KB neural network that dreams ahead to beat you at blocks.

Creator–brand matching that finds the right creator even when their content never mentions your product.

A worker-pool and model-server toolkit for running vision, embedding, and chat-completion LLM workloads, interfaced through code (not HTTP).

Real-time hand gesture recognition controlling a VR scene on Meta Quest 2 — PyTorch · ONNX · C++ · Unity

EuroSAT 6-Band Image Classification using MobileNetV3 with ONNX export

A production-ready framework featuring learnable gates for automated model pruning, achieving 90%+ compression. Supports MobileNetV3, ResNet-18, EfficientNet-B0 with Streamlit dashboard, ONNX export, Docker deployment, and CI/CD. Ideal for AI optimization research and deployment.

This project benchmarks vision (MobileNet, ResNet20) and NLP (DistilBERT) models across server vs. client inference, using backend servers and ONNX-converted local models. It evaluates latency, UX, accuracy, and performance, dedicated to fulfilling the Master’s thesis in AI & ML.

Production-ready MLOps pipeline for Depth-to-PointCloud conversion powered by ONNX Runtime test-driven CI/CD

Free AI Background Remover 2026 ✨ | Unlimited & Open Source

A ~10M parameter onxx Runtime enabled LLM from scratch trained on Tiny Shakespeare on free tier collab GPU.