JobHuntly
An interview notebook: prepare each round, write down what happened, and see what your search says.
- Role
- Product, architecture, backend, front end and deployment
- Stack
- Next.js 16 · React · .NET 8 Web API · Entity Framework Core · PostgreSQL (Neon) · Groq · Fly.io · Vercel
The problem
A job search is a handful of interviews, and each one needs preparation and a debrief while it is fresh. Spreadsheets and Kanban boards track the applications, but nothing helps you walk into the second round knowing what went wrong in the first.
The solution
JobHuntly makes the interview the unit. You prepare a round, write the debrief straight after it, and the app shows reminders for today, the funnel of your search, and a comparison of offers in USD, GTQ and EUR. A one-click demo fills it with sample data so you can see it working.
What it does
- Prepare each interview round and debrief right after it
- A "Today" view of what needs attention, and the funnel of the search
- An offer comparator across USD, GTQ and EUR
- AI (Groq) that extracts a pasted job offer, suggests likely questions and structures a debrief, and is always a draft you confirm
- The app works fully without the AI
- Backend on Fly.io, database on Neon, front end on Vercel; measured cold start about 7 seconds, warm responses 0.7 seconds
Decisions
Keep the backend in C#
A Node API would have been faster to write. I kept .NET 8 because it is what I want to be hired for, and moved from SQL Server to PostgreSQL so that the database could live on a free tier.
AI drafts, people decide
Nothing the model returns is saved without your confirmation, and every feature has a path with no AI at all. The app is useful offline from the model.
Spec first
Principles, a spec with measurable acceptance criteria, a plan with the alternatives I rejected, and a table that maps each criterion to the test that checks it. They are in the repository.