Applied AI

Varun Agrawal  • Peerlist

This project implements a text-conditioned U-Net that segments drywall defects and features using natural language prompts. Given an image and a prompt like "segment crack" or "segment taping area", the model produces pixel-level binary masks for construction quality assurance.

Crack/Taping Segmentation

This project implements a complete pipeline for fine-tuning small LLMs like Llama 3.1 8B Instruct and integrating them with Mem0's hybrid memory system. The system enables personalized AI conversations that remember context across sessions while running efficiently on consumer hardware.