Built to figure
something out.

The problem

Checking restaurant listings for discounts can mean repeatedly scanning the same page. I wanted a quicker way to see the offers already visible there.

What I built

A Firefox extension that scans visible Wolt listings, extracts discount information and presents it in a popup. The work includes DOM scanning, filtering false positives and keeping the interaction straightforward.

What it taught me

Useful automation starts with separating useful information from noise. The project connects a repetitive task to a clearer interface, with local-only behaviour. The source site links to the Firefox add-on below.

The question

What changes when I host my own search tool and control more of its configuration?

What I explored

A local SearXNG setup for experimenting with search aggregation, privacy-focused tooling and self-hosted services. I used it to practise configuration and troubleshoot how the service behaves.

What it taught me

Running a service locally makes reliability, configuration and the user experience much more concrete. This is a personal experiment, so there is no public demo or service to sign into.

Local experiment. No public demo.

The question

How useful is a local AI model, and where do the limits of my own setup show up?

What I explored

A locally hosted DeepSeek model for testing private prompts, responses, local inference and system requirements. The focus is practical experimentation and understanding the tooling.

What it taught me

Local workflows make resource constraints and response quality easier to examine directly. This is a personal setup, not a hosted AI product or a published benchmark.

Local experiment. No public demo.