Marc Shaw

Problem

Finding suitable IT roles across Switzerland meant repeatedly searching job boards, recruiter listings and individual employer websites. Vacancies were duplicated, many results were irrelevant, and good employers could be missed entirely if they advertised only on their own careers page.

Solution

I built a self-hosted job discovery and matching platform that automates hours of manual job searching. It discovers employers, identifies their careers websites, monitors vacancies and evaluates opportunities against my skills and preferences. It also tracks applications, preserves copies of job adverts and sends highly relevant opportunities directly to me via Telegram. The result is a continuously expanding, personalised job search system that saves time and helps ensure promising opportunities aren’t missed.

Business value

The system turns a repetitive manual search into an automated monitoring workflow. Employers remain on the watchlist even when they currently have no vacancies, so suitable roles can be detected when they appear. The same architecture could be adapted for recruitment, talent sourcing or other vacancy-monitoring use cases.

AI-assisted

I defined the matching logic, employer-discovery workflow, validation rules and safety requirements; Cursor was used to implement, test and iteratively improve the application using real vacancy and employer data.

Python · FastAPI · SQLAlchemy · SQLite · Jinja2 · Brave Search API · LINDAS · AI-assisted development

View case study

Problem

Cyber-insurance assessments relied on questionnaires that were generic, hard for non-technical managers to answer accurately, and quickly out of date.

Solution

For my former employer in insurance, I designed and built a platform that assesses a company’s security from public evidence and produces clear broker reports.

Impact

Every finding traces back to its evidence and scores are repeatable, giving insurers, brokers and clients a clearer, more objective basis for decisions.

AI-assisted

AI only turns verified findings into plain-language explanations. It doesn’t create findings or calculate scores, so results stay auditable.

Angular · Go · MongoDB · Auth0 · Docker

View case study

Problem

Hundreds of family photographs spanning around 40 years needed digitising. Feeding old prints through an ADF risked jams or damage, while scanning them individually on a flatbed would take far too long. The available off-the-shelf workflow required separate scanning and paid photo-splitting software.

Solution

I used AI-assisted development to build one desktop application that controls the flatbed scanner, scans several photographs at once, automatically separates them into individual files, and organises them into named archive batches ready to upload directly to my home media centre.

Business value

The tool was built in a day using Cursor, which costs about CHF 20/month, replacing the need for additional commercial software and a multi-application workflow. It can be reused for future digitisation work and could be packaged for reuse or commercial use.

AI-assisted

I defined the problem, workflow and validation requirements; Cursor was used for implementation, debugging, testing and rapid iteration against real scans.

Python · OpenCV · PySide6 · SANE · AI-assisted development

View case study

Problem

Family finances span personal and business accounts, savings, pensions and insurance, plus incoming letters that had to be read, filed and acted on by hand.

Solution

A self-hosted app that turns bank statements into a verified transaction ledger, flags missing statement periods and organises incoming documents.

Impact

One private place to see balances, savings and policies over time. Figures are checked before they’re trusted, and our financial history stays off third-party clouds.

AI-assisted

Rule-based checks protect the ledger. AI is reserved for understanding correspondence — translating and summarising letters and flagging when action is needed — which is still being developed.

Next.js · FastAPI · PostgreSQL · MinIO · Tesseract · Docker

View case study

Problem

Household media, photos and files needed a private home with dependable backups and remote access that isn’t exposed to the public internet.

Solution

A self-hosted platform on Ubuntu and Docker with redundant storage, encrypted backups and VPN-only access, which I designed and run day to day.

Impact

Family data stays private and recoverable, and services can be reached remotely without being published online.

AI-assisted

AI supports operations — documentation, troubleshooting and change planning — while live checks and privileged changes stay with me.

Ubuntu · Docker Compose · Caddy · Tailscale · RAID1 · Restic · LUKS

View case study