Skip to content
Indian Software Developers
Menu

Case study, status: r&d

Local LLM inference and fine-tuning lab

A private setup for running and fine-tuning open-weight models on a single consumer GPU.

Overview

Problem

Some clients cannot send data to third-party model APIs, and API costs grow with volume. Self-hosting needs real knowledge of the trade-offs.

Objective

Find out what quality and speed are realistic on a 16 GB consumer GPU, and build a repeatable fine-tuning workflow.

Solution

  • Runs llama-server with GGUF models on an RTX 5080 16 GB; Q4_K_S quantisation was the practical choice for a 27B model.
  • Compared vLLM on Windows through WSL2 with a community native Blackwell build.
  • Fine-tuned with Unsloth using LoRA and QLoRA, including a Llama 3.2 1B experiment on generated stock OHLCV analysis data.
  • Ported an MLX fine-tuning script from macOS to Windows and CUDA.

Architecture

  1. Client app

  2. OpenAI-compatible endpoint

  3. llama-server / vLLM

  4. Quantised GGUF model

  5. RTX 5080 16 GB

AI workflow

  1. 1Choose model size and quantisation to fit available VRAM
  2. 2Serve through an OpenAI-compatible API
  3. 3Generate a domain dataset for fine-tuning
  4. 4Train LoRA / QLoRA adapters with Unsloth
  5. 5Compare outputs against the base model

Technology

  • llama.cpp (llama-server)
  • GGUF
  • vLLM
  • Unsloth
  • MLX
  • CUDA

Want something similar built?

Send a short brief. We'll reply personally with questions or a suggested next step — usually a technical discovery call.