Back to Dashboard

Project Demos

AI & deep learning projects in action — from LLM fine-tuning tools to real-time vision systems and edge inference.

YouTube Channel 8 Demos
Video Highlights
LLM Tools
Fine-tuning desktop app with torchtune
100+ FPS
Real-time detection with FlashDet
Galaxy S23/S24
Samsung production deployment
Jetson TX2
Edge inference on NVIDIA hardware
Featured LLM Fine-Tuning Desktop App
LLM Tool

LLM Fine-Tuning Desktop App

PyQt5 desktop GUI for fine-tuning LLMs using torchtune — visual workflow for LoRA/QLoRA training, evaluation, and one-click deployment. No command-line required.

torchtuneLoRAQLoRAPyQt5Desktop
Watch on YouTube
All Demos
Detection FlashDet Real-Time Object Detection
FlashDet — Real-Time Object Detection
End-to-end detection system with LoRA fine-tuning, knowledge distillation, ONNX export, and INT8 quantization. 100+ FPS inference.
LoRAKDONNX100+ FPS
GenAI AI Object Removal & Inpainting
AI Object Removal & Inpainting
Remove objects from photos including shadows and reflections using generative inpainting with diffusion-based restoration.
DiffusionInpaintingGenAI
Vision Low-Light Image Enhancement
Low-Light Image Enhancement Pipeline
SOTA deep learning denoising and enhancement for low-light images — mobile camera pipeline deployed on Samsung Galaxy devices.
DenoisingSamsungMobile
Edge AI ALPR on Jetson TX2
ALPR System on Jetson TX2
Automatic License Plate Recognition with anchor-based detector + multi-head OCR running real-time on NVIDIA Jetson TX2.
Jetson TX2OCREdge
Analytics Video Analytics Pipeline
Video Analytics — Traffic & People Counting
Production video analytics system for traffic detection and people counting with multi-object tracking.
TrackingAnalyticsProduction
Quantization Model Quantization Demo
Model Quantization & Mobile Deployment
End-to-end quantization pipeline — GPTQ, QAT, INT8 for efficient model deployment on mobile and edge devices.
QATINT8Mobile
Short AI Quick Demo
AI in Action — Quick Demo
Quick showcase of AI capabilities in a short-form video format.
ShortDemo