FAANG interviews fit into 8 weeks of preparation if you sequence it. Week 0 is finding out the loop from the recruiter.
On this page
Weeks 1 and 2 are the 15 patterns, not 300 problems. Weeks 3 and 4 are system design out loud.
Week 5 is 8 written stories. Week 6 is scored mocks. Week 8 is rest.
My first Google interview lasted 7 minutes before I failed it.
That is still my record. I had not understood the format, so I started coding before I understood the question.
I have since sat in more than 100 interviews, on both sides.
Google, Amazon, Meta, Shopify.
Here is the 8-week plan I would hand that version of me.
Week 0: what are you actually being tested on?
The loop, before you solve a single problem. Ask the recruiter how many rounds, which types, and what level.
They answer this and almost nobody asks.
A Staff loop and a Senior loop share the same questions and score them differently.
Prepare for the wrong one and you pass every round with no offer.
Weeks 1 to 2: how many patterns do you need?
About 15, and they sit behind most interview questions.
Two pointers, sliding window, binary search on the answer.
BFS and DFS, topological sort, heaps, intervals, union find, and a few more.
Work 4 to 6 problems per pattern. Cannot name the pattern within 90 seconds? You are not done.
After each problem, write down what made the pattern applicable. That sentence is what you need under pressure, not the code.
That 90-second test is what I built AlgoPath around.
It teaches the 15 patterns through the places they run, not as a list.
Its trainer hands you an unlabelled problem and asks which pattern it is.
Weeks 3 to 4: why does system design have to be out loud?
The round tests whether you can turn a vague prompt into requirements.
Then pick a shape and defend it when someone pushes.
You already know what a queue is.
Speaking is what collapses under pressure. Practise the same 4 steps out loud, recorded.
The recording is uncomfortable and it is the fastest feedback available.
The 4 steps, every time. Clarify the requirements and write them down. Estimate the scale in numbers. Sketch the shape.
Then go deep on the one part the interviewer cares about.
System design is where senior candidates lose offers most often.
Week 5: behavioural, with real stories
Most candidates improvise here and it shows.
Write down 8 stories from your actual work.
Conflict, failure, influence without authority, a decision you got wrong.
Changing your mind, delivery under pressure, mentoring, disagreeing with a manager.
Each story needs the situation in 2 sentences.
Then what you specifically did, and the outcome with a number.
Interviewers are trained to ask follow-ups. An invented story falls apart on the second question. A real one gets better.
Week 6: run the whole loop under pressure
Do full-length mocks with someone who fills in a scorecard.
Someone who tells you where you landed against a real bar. Not a friend who says it went well.
The gap between practice and performance is what a watching stranger does to your thinking.
Week 7: fix the 2 things that keep breaking
By now you have a pattern of failure. Everyone does.
Maybe you go silent when stuck. Maybe you start coding before clarifying. Maybe you cannot finish in time.
Pick the 2 biggest and drill only those. 1 more pattern will not help you now.
Week 8: taper
Do not cram in the last week.
2 easy problems a day to stay warm. Re-read your 8 stories. Sleep properly.
Walking into the room tired has cost more offers than any missing algorithm.
What gets people rejected in FAANG interviews?
On the hiring side, rejections almost never say the candidate did not know the algorithm.
They say the candidate did not clarify the requirements.
Or jumped to an implementation without explaining why. Or went silent for 4 minutes.
Or could not say what they would do differently.
All of those are fixable in a week.
None of them are on LeetCode.
Does a referral help?
It gets your CV read and it does not lower the bar.
Ask people who have seen your work. A stranger who refers you is guessing, and it shows.
A referral is worth most in week 0, when you decide which loops to run.
Common questions
Is 8 weeks enough to prepare for a FAANG interview?
Enough if you already write code daily and have shipped real systems. Not enough from a standing start, where 4 to 6 months is realistic. The 8 weeks assume roughly 10 hours a week. Compressing the plan into 3 weeks produces recall without judgement. That is what the loop screens out.
How many LeetCode problems should you do?
Around 80, spread across 15 patterns, beats 300 chosen at random. The count that matters is how many patterns you can name on sight. Not how many problems you have seen. Track the naming, not the solving. Stop adding problems once you can name a pattern in 90 seconds.
Should you prepare in your strongest language?
Yes, and pick one and stay in it. The interview measures thinking under pressure. Switching languages adds a tax exactly when you have no room for one. Python and Java are safest for readability in a shared editor. Tell the interviewer which one you are using, and why.
When should you book the interview?
Book it at week 0, for week 9. An unbooked interview lets the preparation drift, and week 8 becomes week 20. Book a lower-priority company first. Run their loop for real, and treat it as the mock nobody can simulate. Then take the loop you want.
The short version
Find out the loop before you prepare. Learn 15 patterns, not 300 problems. Practise system design out loud.
Write your stories down. Get scored by someone real before it counts.
That last part is what a mock interview is for.
7 minutes was a fair result for how prepared I was. It does not have to be yours.