Preparing Your Evidence

Interviews are evidence-gathering exercises, and most candidates show up with the evidence unorganized. You have years of real work; the prep task is mapping it to the questions in advance so nothing has to be composed live.

The evidence map

Take the job description and extract its claims — for a .NET + AI transcript-review role: ASP.NET Core MVC, EF Core, a UI grid stack like Kendo, Azure DevOps pipelines, and AI-assisted analysis with human review. For each claim, write down the specific project, the problem, and one number. "I've used EF Core" is not evidence. "I cut a review-queue page from 9 seconds to 400ms by replacing lazy-loaded navigation walks with a single projection" is evidence — dated, sized, and yours.

Numbers matter more than adjectives: rows migrated, requests per second, minutes saved, percentage of flags reviewed. Approximate honestly ("roughly 100k pages") — precision theater is a red flag, but no numbers reads as no ownership.

The 6–8 stories to have ready

Behavioral and scenario rounds recycle the same territory. Prepare these, each in the situation → action → result → what I'd do differently shape, 90 seconds long:

1. A production bug you owned end-to-end — the debugging story.
2. A performance win — with the before/after number.
3. A design decision with a real trade-off — where you chose, and what it cost.
4. A disagreement — technical conflict, resolved like an adult.
5. A failure — genuine, with the changed behavior that followed. (The non-answer "my weakness is perfectionism" fails the round.)
6. Learning something fast — new stack, tight deadline.
7. An AI/LLM story — see below.
8. Explaining something technical to a non-technical person — gold for public-sector panels.

The portfolio project is your AI evidence

Most MVC developers interviewing for AI-adjacent roles have no production AI story. If you've built anything real — say, an AI learning platform with RAG-based tutoring, LLM-generated content behind human review, and text-to-speech — that is your differentiator, so rehearse it as engineering, not as a demo: what the retrieval pipeline does, how you handled hallucination (grounding answers in course content, constraining the prompt), what you'd add before trusting it at scale. The phrase that lands in a transcript-review context: "I've already built the pattern this job needs — AI drafts, a human decides." A side project discussed with production-grade seriousness beats a buzzword list every time.

Practice prompts:

1. Write the one-number version of your best performance story, then say the 90-second version out loud.
2. Answer "Tell me about a time you used AI in a real system" using your strongest project — include one limitation you hit.
3. For each requirement in a real job posting, name your matching evidence in one sentence. Any requirement with no sentence is your prep gap.