Mock interviews — anatomy of a good practice loop.
The biggest gap in interview prep is practicing actual delivery. You can read a hundred STAR-format articles and still freeze the moment a stranger asks "tell me about a time you failed." This page is the field manual for what a good mock loop looks like — the five-step structure, the rubric a real interviewer is scoring on, the per-company calibration, and the four-week daily-rep schedule. Use it with a peer, a coach, or your own recorder; the loop is what matters.
Contents
- The structure of a good mock session
- Five modes — what each one practices
- The grading rubric — what an interviewer is actually scoring
- Per-company calibration — Amazon vs Meta vs Google
- The four-week daily practice loop
- Tools you can use tonight — honest landscape scan
▸ Just want to start practicing? Skip to §06 — tools.
§ 01 — The session structure
The structure of a good mock session.
Every productive mock — peer, coach, or self-recorded — has the same five-step loop. Each round runs 30-60 minutes depending on the mode.
- Pick a mode — Behavioral / Technical Coding / System Design / Cross-functional / Career Story
- Pick a target — generic FAANG, Amazon LP-driven, Meta behavioral, Google Googliness, OpenAI mission-driven, etc.
- The interviewer asks — one question, clean prompt, no priming
- You respond — uninterrupted; if it's a self-mock, record it so you can play it back
- Grade + follow-up — score the response against the rubric below, then run the bar-raiser drill-in question that targets the weakest signal
Sample session · Amazon LP-driven behavioral
Interviewer: Tell me about a time you had to make a decision with incomplete information. (Targeting: Are Right, A Lot)
You: Last year my team was deciding whether to migrate from Kafka to Kinesis. The vendor cost was 40% higher with Kinesis, but the integration story was cleaner. We had two weeks before the decision had to ship to leadership and only 30% of the data we wanted...
Coach notes: Strong opening — you named the specific decision, the time pressure, the data gap. Three things to tighten: (1) You said "my team" — switch to "I" for ownership signal. (2) Quantify the stakes — what did getting it wrong cost? (3) The Action section needs 2-3 more specific moves. Score: 7.5/10.
Follow-up: What specific data did you have, and what did you decide to extrapolate? Walk me through your reasoning step by step.
· · ·
§ 02 — Five modes
Five modes — pick what to practice.
| Mode | Duration | What it practices | Output |
|---|---|---|---|
| Behavioral | 30-45 min | STAR delivery, story selection, level calibration, follow-up handling | Per-story rubric + transcript with annotations |
| Technical Coding | 45-60 min | Live coding with the AI asking clarifying questions, runtime/space analysis, optimization probes | Code review + complexity analysis + interviewer-style critique |
| System Design | 45-60 min | Whiteboard-style; AI plays "interested interviewer" and pushes on bottlenecks | Diagram critique + missed-trade-off list |
| Cross-functional | 30 min | Stakeholder management stories, conflict resolution, PM/data-scientist partnership signals | Multi-axis rubric scoring |
| Career Story | 45 min | "Tell me about yourself" + the L+1 promotion narrative | Pacing critique + scope-signal check |
· · ·
§ 03 — Grading rubric
The grading rubric — what an interviewer is actually scoring.
Behavioral mode
| Signal | Score 1-10 | What it measures |
|---|---|---|
| Structure (STAR pacing) | 1-10 | Time budget on each phase; clear transitions |
| "I" vs "we" ratio | 1-10 | Personal ownership signal |
| Quantified results | 1-10 | Numbers attached to outcomes |
| Specificity of actions | 1-10 | Concrete decisions vs generic verbs |
| Level calibration | 1-10 | Scope of story matches target level |
| LP / signal mapping | 1-10 | Does the story actually hit the LP it was asked for? |
| Follow-up resilience | 1-10 | How you handle the bar-raiser drilling in |
Technical coding mode
| Signal | What it measures |
|---|---|
| Clarifying questions | Did you ask before writing? How many? |
| Solution sketching | Did you talk through the approach before coding? |
| Correctness | Does the code pass the test cases? |
| Complexity analysis | Time + space — accurate? |
| Edge cases | Empty input, single element, all-same, max scale |
| Code style | Readable variable names, no dead code, idiomatic |
| Optimization narrative | Did you call out the optimization opportunity? |
· · ·
§ 04 — Per-company calibration
Per-company calibration.
Same question, different grading rubric per company. Pick the target before you start, and weight your self-grade accordingly.
| Target | What the rubric weighs heavily |
|---|---|
| Amazon | LP mapping (each story must hit a specific LP); STAR rigour; specificity of "I" actions; second-story-on-demand |
| Meta | Quantified metrics; cross-functional partner named; impact at scope; speed signals |
| Googliness signals (change-mind-on-data, intellectual humility); General Cognitive Ability ambiguity-handling | |
| Netflix | Autonomy + judgment; decisions made without asking permission; culture-memo alignment |
| OpenAI / Anthropic | Mission-aligned reasoning; specificity of why-this-lab; safety/values nuance |
| Stripe | Writing quality (Stripe famously asks for written responses); long-form coherence; trade-off articulation |
· · ·
§ 05 — Daily practice loop
The candidate workflow — daily practice.
- Day 1-7 · story bank — practice each of the 8 prepared stories in Behavioral mode. Multiple takes; iterate based on the rubric.
- Day 8-14 · LP / signal mapping — random LP drills. Pull a signal at random; you have 5 seconds to pick the right story from your bank.
- Day 15-21 · technical reps — Coding + System Design daily, target 30-45 min each.
- Day 22-28 · target calibration — switch to your target company's rubric. Re-grade old answers against the new weights.
- Day 29-30 · full mock loops — 4-5 rounds back-to-back with breaks, mimicking the real onsite. Stamina training.
The practice intensity. Most candidates over-practice solving problems and under-practice telling stories. Aim for at least 1:1 ratio of behavioral vs technical practice in your prep month. The behavioral round is where most engineers fail; it deserves equal time.
· · ·
§ 06 — Tools that exist today
Tools you can use tonight.
Honest scan of the free / no-account / bring-your-own-key landscape as of 2026. The strict "free and no signup" intersection is small — most mock-interview products gate the first session behind a Google sign-in. Each tool below has the catch spelled out so you can pick by what your evening actually allows.
No signup, free, runs in the browser
- ChatGPT — Mock Interview GPT · best raw quality in this tier (GPT under the hood). Text; limited turns without login, voice mode requires a ChatGPT account. Best for behavioral / STAR rehearsal.
- FavTutor AI Mock Interview · explicitly advertises "no registration, no credit card." Behavioral, HR, coding, system design. Text. Feedback is generic but works as a warm-up.
- GoodSpace AI Mock Interview · no signup, voice-capable, 10,000+ role templates. Shallow on technical depth but real interactive practice.
- Eklavvya Free AI Interview · no signup, ~50 roles, instant scoring. Mostly behavioral / HR.
Bring your own API key — open-source, no account
- Interview Agent (source on GitHub) · paste your own OpenAI key into the browser; it never leaves the page. Four modes: Recruiter, Technical, Hiring Manager, Behavioral. The cleanest pointer if you already have an API key — closest thing to an unlimited, private mock surface that exists for free.
Free first session — signup required
Worth knowing about even though they want a Google sign-in. Each has a meaningful free tier:
- FreeInterviewMe · built by ex-Meta engineers Tejal and Simrat. Free, signup required. Worth a look — the founders sat on the same hiring committees the rubric on this page describes.
- Final Round AI · first mock free, voice. Strong behavioral feedback.
- Yoodli · 5 lifetime free roleplays. Excellent delivery / voice critique.
- Pramp · Exponent Practice · interviewing.io · peer-to-peer mocks with real humans, not AI. Best free option for coding / system-design rounds if you'll accept a sign-in.
- Himalayas · GreatFrontEnd · Interview Query · small free tiers, account required.
The honest gap. For data engineering and system design specifically there is no good free-and-no-signup option as of 2026. The cleanest paths are Interview Agent (BYOK) for solo reps, or the peer-mock platforms (Pramp / interviewing.io) if you'll accept a sign-in. Pair either with the Design pillar as your question source and the Company Tracks for per-company rubric weights.
Last reviewed · 2026. Landscape moves quickly — if a link below is dead or the free tier shrank, ping hello@paddyspeaks.com.
· · ·
Where to go next
Pair this page with the curricula it leans on. The loop above is just scaffolding; the substance is in the playbooks for each round.
▸ Open the question bankRelated pages
- Behavioral & Leadership — the curriculum the mock loop practices
- Company Tracks — per-company calibration profiles
- Real Interview Experiences — what to practice based on real loops
· · ·