Systems in production, described honestly.
Each of these is a real system that was built, deployed and used. The interesting part is usually the constraint that shaped the design, so that is what these write-ups lead with.
What follows is Ingot Solutions’ own product work. Case studies from client engagements are not published here — they are shared on request, with the client’s agreement, and named references can be arranged.
Clarivise Scan — phishing training that happens in the inbox
Ingot Solutions product
Security awareness training is a video in November that everybody clicks through at 1.5× speed. The moment somebody is actually willing to learn about phishing is the moment a suspicious email is sitting in front of them and they are unsure. Scan exists to be useful at exactly that moment, and to make itself progressively less necessary.
- On demand User-invoked — nothing is blocked, held or auto-actioned
- Every link Real destination shown, wrappers decoded
- Plain English Why it is suspicious, not just a score
- Manifest V3 extension
- Office.js add-in
- Supabase Edge Functions
- PostgreSQL
- Playwright
- Claude
Clarivise Shield — AI email security for Microsoft 365
Ingot Solutions product
Microsoft 365 filters the obvious spam. What gets through is the well-written stuff: the vendor-payment redirect, the CEO gift-card ask, the invoice that is almost right. Shield adds an AI reading layer behind the standard filter and marks what it finds — deliberately without the ability to lose anyone mail.
- 60s Maximum delay from delivery to verdict
- Zero Legitimate mail that can be lost by design
- 4-tier Verdict model, each with a defined action
- Python / FastAPI
- Azure Container Apps
- Microsoft Graph
- Supabase Edge Functions
- PostgreSQL
- Claude
Something here look like your problem?
The constraints are rarely identical, but the shapes repeat. Describe what you are dealing with and I will tell you how I would build it — the approach, roughly how long it takes, and what to tackle first.