Writing
Notes on models and the making of them
Practical write-ups on retrieval-augmented generation, fine-tuning open-source models on modest hardware, and the computer-vision projects that started it — published on Medium.
- Published on
- Medium
- 01
Medium · 2025 · 6 min read
Guidelines for Fine-Tuning Open-Source LLMs with Limited Computing Resources
Fine-tuning Qwen2.5-Coder-14B on a single Kaggle P100 using Unsloth, 4-bit quantization and LoRA — dataset prep through to inference.
READ → - 02
Medium · 2025 · 4 min read
LightRAG is lite RAG
Where LightRAG sits against a conventional retrieval-augmented generation pipeline, and what makes it the lighter option.
READ → - 03
Medium · 2022 · 7 min read
Object recognition and handling overfitting for CIFAR-10 image dataset
Building an object recognition model on CIFAR-10, and the techniques that keep it from overfitting.
READ → - 04
Medium · 2022 · 6 min read
Introduction to a machine learning project: Handwritten Digit Recognition
Demonstration of a simple machine learning project aiming that a beginner can learn and understand the basic of it and can apply further.
READ →