Work

Mekloop

B2B SaaS for independent vehicle repair shops

Job management, booking-to-invoicing workflow, parts inventory and customer history, for independent garages. Incorporated June 2026.

Beachhead market identified and tested through direct demos with shop owners and mechanics, with a small group of pilot shops working alongside the initial build. The rollout is deliberately slow and controlled: this is a system holding customer and vehicle records, so getting it right on security, data protection and liability matters more than getting it out fast.

I started out working in my father's repair shop. This is software for that customer.

Digital Lab Report

Deterministic, non-diagnostic explanations of lab results

A consumer application that takes lab test results the user enters manually and returns plain-language explanations — what each value measures, and where the result sits against its reference range. It does not diagnose, and it isn't built to.

Lab reports are written for clinicians. Patients receive them anyway. The gap between "your ALT is 62" and knowing what that means is currently filled by search engines and forums, which is worse than filling it with something bounded and deterministic.

Decisions that constrain the build

• Deterministic, not generative. Explanations resolve from a fixed rule set against reference ranges. No model in the explanation path. A model that is occasionally wrong about a liver enzyme is not acceptable at any confidence threshold.
• Compliance-first. GDPR consent flow and encryption were in from the first commit rather than deferred to a hardening pass. Health data is the wrong category to retrofit.
• Non-diagnostic by construction. There is no code path that produces a diagnosis, so scope creep can't push it into one.

Deliberately excluded from v1

• PDF and image parsing of lab reports
• Clinician-facing features
• Longitudinal trend analysis

Each is a reasonable v2. None is needed to test whether the explanation layer is any good.