manupanand

A personal project where I fine-tuned **Qwen-2.5-7B** using **Unsloth + QLoRA** on an **AWS g6.xlarge instance** (NVIDIA L4, 24 GB VRAM). The goal was to learn and implement memory-efficient LLM fine-tuning end-to-end on a single GPU — from environment setup to training to exporting the final model.

my repo of codes with learning llm rags etc

This is my Bio readme

No project description available.

my repo to lean mesasge queue

A production-grade reference implementation demonstrating how to securely deploy code from a cloud control plane (GCP) to on-premise infrastructure with **zero inbound connectivity, zero standing credentials, and zero VPN** — using OIDC, Workload Identity Federation, and pull-based Ansible automation.