Project : TandemView

A mock interview is only useful if the questions are yours. TandemView reads the job description and the CV you paste in, checks they belong together, and builds every question from them.

Five levels, rising in difficulty, each applying a different prompt-engineering technique. Five roles, from AI engineer to product designer, each with its own interviewer knowledge. You get question cards, graded feedback, hints, and a summary at the end.

The safety work is the part I am proudest of: a judge reads every input before the main model does, user text is wrapped so injected instructions stay inert, and nineteen attack attempts are logged in the repo. The interview quality was scored by an LLM judge across fifty sessions, and the results are committed as evidence rather than claimed.

An interview coach that cannot be talked out of its own rules is the minimum, not the achievement.

What it is made of

An IT mock-interview coach, built in 2026 during the Turing College AI Engineering programme with LangGraph and LangChain, evaluated rather than eyeballed: the sprint shipped with an evaluation set, not just a demo.