Bidirectional fine-tuning of Microsoft's Phi-3-Mini model for payment transaction processing using LoRA. Includes forward (structured→NL) and reverse (NL→structured) models. Optimized for NVIDIA RTX 3060 (12GB VRAM). 500 synthetic examples, ~95% accuracy, 30-60min training time.

Repository for QA-based event detection and extraction from news and social media.

A full-stack invoice management platform built with React, Node.js, and MongoDB that uses Google Gemini to generate invoices from text prompts and track revenue insights.

💡 Fine-tune Microsoft's Phi-3-Mini model for payment processing, converting structured data to natural language and vice versa with ease.