Portfolio
Selected Projects
Below are a few representative AI systems and research projects that demonstrate production readiness, model optimization, and real-time deployment.
DeepVoiceGuard
AI-powered audio spoofing detection for speaker verification systems. Combines RawNet2 audio modeling with FastAPI deployment for real-time inference.
- ASVspoof 2019 dataset
- Audio anti-spoofing and security monitoring
- GitHub
VoiceGuard
Voice phishing detection solution that analyzes speech content and metadata to identify fraudulent calls in real time.
- Real-time classification pipeline
- Fast inference for streaming audio
- GitHub
Face Segmentation
Semantic segmentation of facial features for AR, effects, and image editing workflows using PyTorch.
- Face region segmentation for makeup and AR effects
- Model optimization for inference speed
- GitHub
License Plate Detection (ONNX)
YOLOv5-based license plate detection with ONNX Runtime acceleration for image, video, and webcam input.
- Edge-ready inference with ONNX
- Supports GPU and CPU deployment
- GitHub
Uzbek Sign Language Recognition
Real-time gesture recognition system for Uzbek sign language, built to improve accessibility and communication.
- Mediapipe + OpenCV pipeline
- Real-time text conversion from hand gestures
- GitHub
Explore more projects and research work on GitHub.