Research coordinated into
interview strategy.

Preparing for interviews means connecting company context, role requirements, and evidence from a resume. This multi-step workflow brings those inputs together into a reviewable interview plan.

2-minute screen recording of the real-run replay, specialist outputs, architecture, and evaluation. Silent, with English captions. The vacancy is a demonstration brief.

From research to an interview plan.

19.0s · gpt-4.1-mini

Real run of the BYOK workflow. Topify AI research + a clearly hypothetical AI Product Engineer brief + selected public resume evidence. Recorded . Playback makes no API calls.

0.0 / 19.0s

User output

Company: Topify AI — https://topify.ai/ Job description and resume evidence supplied.

Download the recorded run (JSON) →

Run an interview-preparation workflow.

Your key goes directly from this browser to OpenAI. This website never receives or stores it. OpenAI API charges apply to your account; a run makes six model calls and one web search. Documents are sent to OpenAI; avoid confidential or sensitive information. The key is cleared from the form when a run starts.

A planner with explicit specialist stages.

  1. 01

    User

    Provides the target role and career context.

  2. 02

    Triage

    Identifies the request scope before deeper work begins.

  3. 03

    Planner

    Breaks the task into company and job-analysis tracks.

  4. 04A

    Company Research

    Builds relevant company context.

  5. 04B

    Job Analyzer

    Extracts role requirements and signals.

  6. 05

    Resume Analyzer

    Maps resume evidence to the role.

  7. 06

    Strategy Agent

    Synthesizes research and evidence into a strategy.

  8. 07

    Interview Plan

    Produces a human-reviewable preparation plan.

Implementation details.

01Planner-led orchestration
A planning brief precedes two specialist tracks; fixed typed handoffs combine their outputs.
02Specialist boundaries
Company research, job analysis, resume evidence, and strategy remain explicit responsibilities.
03Typed handoffs
Typed stage events carry status, timing, and reviewable text outputs between explicit handoffs.
04Tools + evidence
Company research uses web search and preserves official source URLs for review; offline checks expose coverage and boundary failures.
05Tracing
The BYOK workflow shows stage status and reviewable specialist outputs; it does not expose hidden model reasoning.
06Human review
The system produces inspectable guidance rather than hiding decisions behind one opaque response.

Check the evidence behind the plan.

A company summary should cover the requested facts, support each claim with a source, and identify missing evidence. Resume mapping should distinguish verified experience from gaps.

Company facts
Compare claims with a human-written reference built from official company sources; mark supported, unsupported, or missing.
Role coverage
Check whether the summary and interview plan address each requirement in the supplied job description.
Resume grounding
Trace each experience claim to supplied resume evidence, without inferring credentials.

Recorded-run checks: 7 / 8 passed.

Deterministic checks against a small reference from Topify’s official homepage. These check coverage and handoff boundaries; they are not an accuracy benchmark or blind human evaluation.

CheckRecorded result
Original role + hypothetical-vacancy labelPass
Explicit requirements separated from hypothesesPass
Official company source URLs presentPass
GEO product reference coveredPass
Tracked-engine reference coveredPass
Job technology not asserted as company stackPass
Unverified headquarters claim excludedNeeds review
Resume evidence gaps are surfacedPass

Observed failures: the initial run drifted from the role to a stage name; a second run inferred company technologies from the job brief. The current run separates those inputs, but its Silicon Valley location claim remains unsupported by the checked official reference. Source URLs alone do not establish faithfulness.

Preparation support, with human review.

This is a personal engineering demo with fixed stages, not an autonomous hiring or application system. The recorded vacancy is hypothetical. Web evidence may be incomplete or stale, and model outputs can contain unsupported claims. Review sources and resume mappings before using the plan. Runs have a three-minute deadline and no persistent memory. OpenAI charges apply to BYOK runs.

Evidence in. Reviewable strategy out.

The final interview plan remains connected to the research and resume evidence that produced it.