Mock interviews and skill plans · Runs in your browser
Build the skills. Then build for the interview.
For data, cloud and AI engineers. Whichever of these two you need, it starts free and ends in a plan with days on it.
If you have a date
Build for the interview
Six rounds generated around the job description you paste in, from behavioural to finding the bug, each one scored with the reasoning laid out.
- Technical round free, no account
- Paste a JD and the questions follow it
- An Interview Plan up to 45 days out
If you have a gap
Build the skills
12 adaptive questions, each spent where your answer is least predictable, place you on the part of the 166-skill map they reach — then the plan is built from the things you actually got wrong.
- Assessment and skill map free
- 166 skills, 15 areas, real prerequisites
- A Skill Plan up to 180 days out
Already know what you need? Interview Plan · Skill Plan
Runs privately in your browser · No account needed · ₹199 per month, cancel anytime
One scoreboard under both
Whichever door you pick, it is the same 166 skills being measured.
Every round, lab and course evaluation you finish records which skills its questions were actually about. So interview practice moves your skill map, skilling up changes what the Interview Plan drills, and neither half makes you start over. You can run an Interview Plan and a Skill Plan at the same time and they keep separate progress on the same evidence.
One subscription covers the plans on both sides. The assessment, the skill map and the first technical round are free either way.
Half one · Build for the interview
You have a date. Six rounds before it.
Everything from here to the skills half is the interview side: what the rounds are, why a chatbot is not the same thing, and the dated plan that gets you through them.
The interview path
Three steps, about ten minutes
This is the half for a date in the diary. The skilling half is further down, and starts somewhere else entirely.
Paste the JD
Drop in the job description for the role you are actually going for. It never leaves your browser.
Pick a round
Six of them, from technical judgement to finding the bug. Each one is generated around that job description, not pulled from a generic bank.
Get scored feedback
A score, the reasoning behind every answer you got wrong, and a study plan that reorders itself around whatever you keep missing.
The full loop
Six rounds. One of them is free.
Not just the two rounds that are easy to build. Every one is generated around your target role and the job description you paste in, every one is scored, and what you get wrong feeds straight back into your study plan.
Technical Scenario
Real-world judgement calls for your target role. Timed, scored, with the reasoning laid out after every answer.
10 minutes · no accountBehavioural
The questions that decide most interviews. Judged on structure and specifics rather than on a right answer.
STAR, block by blockSystem Design
Design under time pressure, then get probed on the parts you left vague, the way a real interviewer would.
Diagram and trade-offsFind the Bug
Read code that already runs and say what is wrong with it. The thing you do every week and never practise.
One failure class a caseCoding
Spot the trap before the test suite does: the off-by-one, the silent mutation, the loop that returns too early.
Python, Java or JavaScriptAlgorithms
Pick the right approach and defend what it costs, which is the part an interviewer actually pushes on.
Reasoning, not typingNo account, no card. The other five unlock with Pro.
The fair question
Why not just ask a chatbot?
You can, and for explaining a concept you should. The difference is what happens between sessions: a chatbot answers, forgets, and answers again. Interview prep is not a question, it is a campaign against your own weak spots over the weeks you have left.
| Asking a chatbot | Skiiling | |
|---|---|---|
| Remembers what you got wrong | A new conversation starts from nothing. You have to remember your own weak spots and bring them up. | Every graded answer is recorded on your device and scored across 166 skills in 15 areas. Evidence fades on purpose, so a skill you have not touched in a fortnight has to be earned again. |
| Tells you what to do next | Answers the question you asked. It cannot know what you should have asked. | A dated plan, reordered from your own results, with one line of evidence for each change — "moved join fan-out earlier, 2 misses last round". |
| Labs you actually run | Describes the bug in prose. You never touch the data, and a plausible-sounding explanation is indistinguishable from a right one. | 75 labs with real tables and a wrong number on a dashboard. You write the query, and it is checked against the right answer. |
| Rounds that are scored | No mark, no pass line, no record. Ask again tomorrow and it starts over. | Six round types over 411 questions, graded per item, each one tagged with the skills it tests (542 tagged so far). |
| Checked before you see it | Generated as you read it. Nothing verified it, including the model. | Every lab and round runs against its own data on every change: 486 checked, 0 failing as of 2026-10-08. Each item carries a visible status, and the method is published. |
| Private by design | Your job description and your answers go to a server, and with most providers may be retained or used to improve the model. | Runs in your browser. There is no account, and the job description you paste and the answers you write never leave your device or enter your progress log — so they cannot reach a backup or a synced copy either. |
Where a chatbot wins: anything open-ended, anything outside these roles, and explaining something five different ways until one lands. Use both. Skiiling is the part that keeps score.
The plan · free to build
You don’t have three months.
You have until the 14th.
Tell it your role, your level and the date — an interview, or just the day you want to be good at this. It works backwards and lays out every day between now and then — which round to rehearse, which lab to break and fix, which pages to read — sized to the time you actually have, and finishing the day before so you arrive rested rather than cramming.
- ✓Built around the date, not a generic 30-day courseThree weeks gets a three-week plan. Five days gets an honest five-day one.
- ✓Every round, every lab, every guide — already yoursIt schedules the Skiiling library you already pay for. Nothing extra to buy.
- ✓Progress tracks itselfFinish a round or a lab anywhere on the site and the plan ticks it off.
- ✓Tailors to the actual job descriptionA model on your own device reads the JD and reorders the plan. The JD never leaves your machine.
18
days until your interview
62
minutes a day, not four hours
6
rounds, rehearsed before the real one
Half two · Build the skills
No date. A gap you keep getting caught by.
The same engine pointed at a different problem: find out what you are actually missing, then work forwards to a date you choose rather than backwards from one you were given.
Build your skills
You cannot close a gap you cannot name.
Most study plans are a reading list someone else wrote. This one starts by finding out what you get wrong, and it is rebuilt every time that changes.
Get measured, not graded
12 scenario questions, each one picked by how you answered the last. It finds the gaps instead of confirming what you already know, and it is free with no account.
Free · about 10 minutesSee where you actually are
Your position on every skill it touched, grouped into 15 areas with the prerequisites between them. Weak areas come with the arithmetic behind them, so you can argue with it.
Free · 166 skillsSay when you need to be ready
A number of days, up to 180. The plan is built backwards from that date, so a three-week run and a six-month run are genuinely different plans rather than the same list at a different pace.
Up to 180 daysKeep missing it and the plan changes
Every finished item is evidence. Miss something twice and it moves earlier and comes back more often; pass it and the plan stops spending your days on it.
Re-planned, not re-orderedWhat the plan is built out of
All 24 certification exams are measured against the same skills, so passing one moves your map. Scheduling them into a plan is not built yet.
The assessment and the skill map are free. The dated plan is part of Pro.
The part that compounds
It finds your weak areas, then keeps coming back to them
A static syllabus cannot know what you keep getting wrong. Every round you finish writes down its own verdict, and the plan and the questions both move.
The round names what went wrong
In its own vocabulary, about itself. Which areas are weak is arithmetic over your own log, not a model’s opinion, so it is the same answer every time.
- the Result blockweak
- data integrityweak
- observabilityweak
Your plan reorders itself
After 2 rounds the material for those areas moves earlier. The plan keeps its length and anything you already finished stays finished.
- 📖 Idempotency and replay
- 🧪 Join fan-out
- 📖 Transactions and isolation
The next round aims there
More of the weak area comes back, and the round tells you so before it starts. A bias, never a filter: the rest of the round still runs.
Before the roundWeighted toward data integrity and security, from your last rounds. The rest of the round is unchanged.
You have to earn it off
Getting it right counts against getting it wrong, and evidence halves every 14 days. That is how the loop closes instead of asking the same question forever.
Your results never leave the browser. The plan is rebuilt on your machine, from your own log.
The whole product in twenty seconds
It gets harder in the places you keep getting wrong.
Four steps, and the fourth one feeds the second. That is the part a chatbot cannot do and a question bank does not try to: the next thing you are asked is chosen by what you got wrong the last time.
You paste the job description
Senior Data Engineer
Spark on EMR, Airflow, a Snowflake warehouse finance closes the books against.
Nothing you paste is stored or uploaded.
You answer a round
Two dashboards report different revenue
Both read the same table. Which number is right, and why?
Question 4 of 10 · find the bug
It grades what you missed
Inner, outer and anti joins
2 misses of 2, last seen 5 days ago.
Worked out on this device, never sent anywhere.
Your plan moves
Moved join fan-out earlier
2 misses last round. Thursday: one lab, 20 minutes.
Nothing you already finished is undone.
Step four hands back to step two, for as many weeks as you have. Start with the free round and the loop begins on your own results.
One subscription, four things
Pro unlocks Role Prep, Practical Labs, E-Courses and Skill Plans on the web.
Role Prep
Six interview rounds generated around your target role and the job description you paste in, each one scored.
Practical Labs
Hands-on debugging challenges you solve right in the page. Broken pipelines, failing queries, real stack traces.
Skill Plans
A dated plan to close the gaps the assessment found, up to 180 days out, rebuilt whenever the evidence changes. The assessment and the map stay free.
E-Courses
Full written course guides to read in the browser, each with a 25-question evaluation and a mock interview of its own.
What this actually is
A small, independent product
Skiiling is built and run by an architect who has been through these interview processes, understands them end to end, and designed the platform around what it actually takes to get through them. The practice runs in your browser, needs no account, and one complete round is free, so you can judge it for yourself rather than take anyone's word for it. Everything below is countable, and you can click through to any of it before paying.
Interview practice covers 12 lab tracks and 14 target roles. Browse the guides or the labs before you decide.
Also available
Skiiling for Mac
Everything above, plus 92 courses, an AI tutor and on-device certifications, running privately on your Mac with Apple Intelligence. A separate purchase from the web subscription.
Not on a Mac? Everything on this page works in your browser. Browse the course catalogue.
Questions people actually ask
What is the difference between the two paths?
The interview path assumes a date: you paste a job description, practise the six rounds you will face, and the plan works backwards from the interview. The skills path assumes a gap: a free assessment places you on every skill this measures, you say how many days you have, and the plan is built from what you actually got wrong. Both can run at once, and both keep score against the same skills, so work on either side moves the other.
Is the skill assessment free?
Yes. The 12-question assessment and the skill map it produces are free with no account, the same way the technical interview round is. The dated plan built from them is part of Pro.
Do I need an account?
No. There is no sign-up, no password and no profile. The free interview runs immediately, and a subscription is unlocked by a licence key rather than a login.
Is it really private?
The job description you paste and the answers you write stay in your browser. Nothing about your practice is uploaded to us or stored on a server.
Do I need the Mac app?
No. Role Prep, Practical Labs, E-Courses and Skill Plans all run on the web with a Pro subscription. The Mac app is a separate purchase that adds the full course library and an AI tutor running on-device.
What's the refund policy?
There is a 30-day money-back guarantee. If you are not happy for any reason, email us within 30 days of your payment and we refund you in full, no questions asked.
Which roles are supported?
Fourteen target roles today, including Data Engineer, ML Engineer, Backend Engineer, DevOps Engineer, Cloud Architect, Data Scientist, Platform Engineer and MLOps Engineer. Pasting a job description tailors the questions further.
Does it work on Windows?
Yes. Everything on this page runs in any modern browser on Windows, Linux, macOS, Android and iOS. Only the separate Skiiling Mac app is macOS only.
Start with whichever one is true.
An interview to get right, or a gap to close. Both start free: no account, no card.