Building a Tiny Guestbook App with Rust and Scratch Containers
Building a small statically linked rust webserver for a guestbook app into a tiny multi-arch docker container.
Solutions Architect & ML Engineer specialising in Information Retrieval
I am a solutions architect with experience building AI-powered systems that solve complex real-world problems. I have a diverse background ranging from simulation engines for mining operations to creating real-time multimodal personalization systems that serve millions of users daily.
Currently, my focus is in information retrieval and vector search technologies. At Marqo, I built scalable multimodal search systems that power major eCommerce platforms, combining cutting-edge research in dense retrieval and multimodal representation learning with production-grade engineering. I actively contribute to the field through research on approximate nearest neighbor algorithms and vector search implementation.
On this site you will find some of my articles, projects, and papers!

Building a small statically linked rust webserver for a guestbook app into a tiny multi-arch docker container.

Creating a small docker container for an updated unbound.

Demonstrating UI components to deliver interesting vector search experiences.


A high performance scalable wrapper around NVIDIA Triton to serve Timm, OpenCLIP and Sentence Transformers models
A zero dependency statically linked Vector Search Engine built with HNSWLib, it supports filtering on schemaless metadata via its own DSL.
A number of caching algorithm implementations written in rust using inner mutability for thread safety. Most interestingly they use the same trait as a base making them modular.
A lightweight, self-hosted guestbook app that lets guests leave star ratings and reviews. Built with Rust and SQLite, the Docker image compresses down to just a few MB.
A lightweight cross platform app for sending HTTP requests and checking responses. Fully local with an included SQLite DB, syntax highlighting, and support for saving and tagging requests.
A dockerised version of Unbound providing recursive DNS resolution. Built from source with the latest security patches, compressing down to just 7MB.