Writing

Everything I've written

Notes from the messy part of the pipeline — reproductions, teardowns, and the things that only show up once you build the thing.

I Built a Multimodal Embedding Model From Scratch on an RTX 4060 (Text, Image, Audio, and Video…

> published: August 22, 2026

I Built a Multimodal Embedding Model From Scratch on an RTX 4060 (Text, Image, Audio, and Video, One Vector Space, 8GB VRAM) How I reproduced the architecture…

How I implemented Self-Adapting Language Models with LoRA and QLoRA on an RTX 4060

> published: July 25, 2026

Self-Adapting Language Models (SEAL): How I Taught an LLM to Teach Itself Implement self-improving LLMs locally. A complete guide to ReST-EM, LoRA, and 4-bit…

How to Build a Vision Language Model from Scratch Using Q-Former, Contrastive Learning, and LoRA

> published: June 27, 2026

Train an image captioning and visual question answering model on 50,000 images in under 4 hours on a single GPU. This post is a deep dive into the open source…

Spread-Spectrum Watermarking: The Invisible Signature in Your Audio & Video

> published: June 20, 2026

How your favorite Netflix show secretly carries a hidden fingerprint and how it catches pirates. The Plot Twist Imagine you’re a music producer. You send your…

Giving Claude Eyes: Building a World Aware Agent with Model Context Protocol and Local Vision…

> published: May 7, 2026

Giving Claude Eyes: Building a World Aware Agent with Model Context Protocol and Local Vision Models In the current landscape of artificial intelligence, we…

I Built a License Plate Detector That Runs on a Raspberry Pi (No Ultralytics License Required!)

> published: March 21, 2026

Ever looked at a license plate detector demo and thought, “Cool, but I can’t afford an enterprise license” ? Same. So I built my own — from scratch, using…

Ditching the Bloat: Building a Native Keras YOLOv8 Edge Object Detector (TFLite & CoreML)

> published: March 18, 2026

If you have ever tried to deploy an object detection model to a mobile phone or a Raspberry Pi, you know the pain. You train a massive, state-of-the-art model…

Quantizing OWLv2 for Efficient Open Vocabulary Object Detection

> published: February 25, 2026

Open vocabulary object detection models allow detection of arbitrary objects using natural language prompts instead of fixed training classes. Among these…

Fine-Tuning OWL-ViT on a Custom Object Detection Dataset

> published: February 25, 2026

Fine-tuning OWL-ViT on a custom dataset requires building a complete detection pipeline because OWL-ViT is fundamentally different from traditional detectors…

Fine-Tuning MantraNet for Image Forgery Detection

> published: August 24, 2025

Cracking Image Forgeries with ManTraNet: Clear Fine‑Tuning Paths, Practical Code, and Smarter Results ManTraNet is a deep learning model for image forgery…