Early access — InterviewCraft is actively in development. New interview types and professions are being added. Expect improvements every week.

Deliberate Practice Engine for Tech Interviews

Stop practicing blindly.
See exactly what to fix.

AI voice interviewer that scores every answer with evidence-backed feedback, rewrites your answer three ways, tracks 22 microskills over time, and builds a personalized drill plan.

No credit card required. 2 free sessions included — then add your own API keys to continue.

The Closed Training Loop

1Answer by voice
2Lint with evidence
3See 3 rewrites
4Rewind + re-answer
5Delta score shown
6Skill graph updates
7Drill plan adapts
8Next weakness drilled

Built Different

Real-time Voice

VAD-aware pipeline with Deepgram Nova-2 + ElevenLabs. No typing. Target p95 < 2s.

Evidence-backed Scoring

15-rule rubric. Every triggered rule links to the exact moment you said it.

Rewind Micro-Practice

Re-answer weak questions. See the delta: +12 in structure, -3 in depth.

Skill Graph Memory

22 microskills tracked across every session. Spaced repetition for weak areas.

Story Bank

Auto-detect STAR stories. Coverage map shows gaps. Overuse warnings.

Negotiation Simulator

AI recruiter with hidden max budget. Scores anchoring and emotional control.

Transparent Pricing

Free to start

Every new account includes 2 complete interview sessions at no cost. No credit card. No hidden charges.

  • Full voice pipeline with scoring
  • Skill graph and drill plan
  • All session types included
Bring Your Own Keys

After your 2 free sessions, add your own API keys in Settings → API Keys. You pay the AI providers directly at their published rates — typically ~$0.10–0.40 per session.

  • Anthropic Claude (voice + scoring)
  • Deepgram (speech-to-text)
  • ElevenLabs (text-to-speech, optional)

Your API keys are encrypted at rest and never logged or shared. You can delete them at any time.

Built for senior technical interviews

Every session type has its own scoring rubric, question bank, and company playbook. The skill graph tracks 22 microskills across all session types — weaknesses surface regardless of which format exposed them.

Behavioral

STAR · leadership · conflict · growth

Structured storytelling, quantifiable results, ownership signal

Playbooks: Google, Meta, Amazon, Stripe

System Design

SWE · Staff · Principal

Capacity estimation, tradeoff analysis, API design, reliability

Playbooks: Uber, Airbnb, Datadog, Cloudflare

Coding Discussion

Technical depth · no whiteboard

Complexity analysis, debugging approach, testing strategy

Playbooks: All engineering orgs

Salary Negotiation

Offers · counter · equity

Anchoring, value articulation, counter-strategy, emotional control

Playbooks: Any offer stage

Debrief

Post-session · coach mode

Pattern recognition across sessions, targeted drill assignment

Diagnostic

Baseline · first session

Full 22-skill assessment to seed the skill graph

Built in public

The full source — backend, frontend, CI pipeline, benchmark suite, and the multi-agent Claude Code development workflow — is on GitHub. Architecture decisions are documented in ADRs. Scoring quality, voice latency, and memory recall are tracked with benchmarks, not vibes.

940ms

p95 voice latency

0.91

scoring Pearson r

4.1 pts

avg scoring variance

View source on GitHub →

Benchmark Results

Measured, not estimated

940ms

p95 voice latency

4.1 pts

scoring variance

0.91

scoring Pearson r

95%

memory recall

Ready to close the loop?

Your first diagnostic session is 5 minutes. It seeds your skill graph and tells you exactly what to practice next.

2 sessions free. No credit card. After that, add your own API keys in Settings.

Start your first session free