Coding interview prep for software engineers
Practice data structures and algorithms, object-oriented design, SQL and system design in Java or Python. Your code runs against hidden tests, and AI Proof Engineers keeps track of what you actually know, skill by skill, so you always know what to practice next.
Dashboard · DSA
Where you stand
Next up: Dynamic programming. You keep missing the base case on 1D problems, so start there.
class Solution {
int search(int[] nums, int target) {
int lo = 0, hi = nums.length - 1;
while (lo <= hi) {
int mid = lo + (hi - lo) / 2;
if (nums[mid] == target) return mid;
if (nums[mid] < target) lo = mid + 1;
else hi = mid - 1;
}
return -1;
}
}A paid ticket never frees its spotFound · line 14
getSpots leaks the internal listMissed · here's why it matters
Why AI Proof Engineers
"Solved" is a weak signal. So is a chatbot telling you your solution "looks correct." AI Proof Engineers gives you something you can check: tests that run, results that match, and a picture of your skills that updates every time you're graded.
Java and Python run against hidden tests in locked-down containers. SQL results are compared row by row in PostgreSQL 18. System designs pass or fail clear rules. No AI decides whether you're right.
Every graded attempt updates your mastery of that skill. Each wrong answer is tagged with the misconception it reveals, so the feedback names the gap instead of just marking you wrong.
Interviews aren't only algorithm puzzles, and neither is the job. Review a pull request, fix a bug in a shipped service, write the tests, tune a slow query, and defend an architecture on a canvas.
Try one
A real question from the SQL course. Pick an answer: a wrong one tells you which idea it comes from, not just that it's wrong.
SQL · free question
customers
| id | name | city |
|---|---|---|
| 1 | Asha | Austin |
| 2 | Ben | NULL |
| 3 | Chen | Dublin |
| 4 | Dana | Austin |
orders (order 14 is a guest checkout, with no customer)
| id | customer_id | total | status | created_at |
|---|---|---|---|---|
| 10 | 1 | 250 | paid | 2026-01-05 10:00 |
| 11 | 1 | 100 | refunded | 2026-01-31 18:30 |
| 12 | 3 | 400 | paid | 2026-01-20 09:15 |
| 13 | 3 | 50 | shipped | 2026-02-01 00:00 |
| 14 | NULL | 75 | paid | 2026-01-31 23:59 |
SELECT name FROM customers
WHERE id NOT IN (SELECT customer_id FROM orders);What does it return?
The subquery returns 1, 1, 3, 3 and NULL (guest order 14). id NOT IN (..., NULL) means id <> 1 AND id <> 3 AND id <> NULL, and the last comparison is unknown - so the condition is never true, for any customer. No error, just an empty result. NOT EXISTS (SELECT 1 FROM orders o WHERE o.customer_id = c.id) returns the intended Ben and Dana, and so does adding WHERE customer_id IS NOT NULL inside the subquery.
No rows at all.
The subquery returns 1, 1, 3, 3 and NULL (guest order 14). id NOT IN (..., NULL) means id <> 1 AND id <> 3 AND id <> NULL, and the last comparison is unknown - so the condition is never true, for any customer. No error, just an empty result. NOT EXISTS (SELECT 1 FROM orders o WHERE o.customer_id = c.id) returns the intended Ben and Dana, and so does adding WHERE customer_id IS NOT NULL inside the subquery.
The courses
Written for the way US, UK and EU tech companies interview. Every problem is checked against a reference solution in the real sandbox before it ships, and each skill has a lesson, a staff-level deep dive, a quick revise page and practice that's graded automatically.
Arrays to dynamic programming, pattern by pattern. Every problem has a hint ladder, visible examples and a hidden suite, in Java or Python.
SOLID and the classic design patterns, then multi-file projects (Parking Lot, Splitwise, a rate limiter) where a hidden test suite decides whether your design holds up.
PostgreSQL 18 in a sandbox. Results are compared row by row against hidden datasets, and tuning labs grade the query plan you produce.
Draw the architecture on a canvas and submit it. Clear rules catch the missing load balancer, the single point of failure and the cache you never justified.
The language rules interviews probe: collections, equality, threads and concurrency, asyncio, and what each release added, from Java 8 to 21 and Python 3.8 to 3.14.
Skills aren't silos. A shaky grasp of recursion shows up in trees, backtracking and DP. AI Proof Engineers rolls the evidence up, so your weak-area report points at the real gap, not the last topic you happened to open.
See how mastery worksWho it's for
New grad or career switcher
Learn the DSA patterns that come up again and again, with lessons that start simple. The hint ladder keeps you moving without handing over the answer.
Start with DSAMid-level backend engineer
Senior loops test design, not just puzzles. Build object-oriented projects, review and debug real code, write production SQL, and design systems that survive a failure.
Start with system designSenior or staff engineer
Staff-level deep dives on internals, costs at scale and failure modes. Concurrency in Java and Python, query plans in PostgreSQL, and what each language version changed.
Start with Java & PythonHow it works
A loop built around evidence, not watching videos. Every step feeds the next.
Each skill opens with a lesson that goes from simple to technical, with diagrams and code in Java and Python tabs. A staff-level deep dive goes further, and a one-page quick revise is there the night before an interview.
Lesson · Deep dive · Quick reviseWrite the solution yourself, then Run it on the examples or Submit it to the full hidden suite. Stuck? Unlock the hint ladder one step at a time; each step you use lowers what the solve earns, so your evidence stays honest.
Run · Submit · Four-step hintsEach skill gets a review date. Recap shows what's due today, weak concepts become a targeted assignment, and the dashboard tells you what to do next and why.
Recap · Weak areas · What to do nextEvery format in one place
The mastery model
A percentage is easy to game and hard to trust. AI Proof Engineers shows the evidence behind every claim: how many graded attempts you've made, how recently, and which concepts keep tripping you up.
Not enough evidence, Developing, Moderate or Strong, with the attempts behind the label.
Spaced repetition brings a skill back sooner when your evidence is thin and later when it's solid.
Course, topic, concept, weakest first, with the questions you recently missed.
Seven competencies, from problem solving to code review, weighted for your target role, show the two gaps worth closing next.
Progress · weak areas
8 attempts · duplicates not skipped
12 attempts · iterative vs recursive order
7 attempts · stale reads after writes
9 attempts · ROWS vs RANGE boundaries
"Sorting the input is enough to avoid duplicate subsets." You picked this on Subsets II. Sorting only puts duplicates side by side; you still have to skip them at the same depth.
The difference
Most prep sites count what you've solved. AI Proof Engineers measures what you know. Here's the difference, side by side.
| What matters | AI Proof Engineers | Typical prep site |
|---|---|---|
| Correctness | Hidden tests, row-by-row SQL, rule-checked designs | Hidden tests for coding; design and SQL often self-graded |
| Feedback | The concept you missed and the misconception behind it | "Wrong answer. Try again." |
| Progress | Evidence per skill, spaced review, weak-area report | A streak and a solved count |
| Formats | Coding, projects, debugging, code review, test writing, SQL tuning, design canvas | Mostly algorithm puzzles |
| Languages | Java and Python, in lessons and every code format | Many languages for puzzles; lessons in one |
| Hints | Four steps: nudge, pattern, approach, skeleton | The full solution, or a chatbot |
| Free plan | Every lesson and question, plus graded samples of every format | A slice of the problem list |
"Typical prep site" describes a common pattern, not one product. Several sites do parts of this well; AI Proof Engineers's difference is doing all of it on one skill graph.
Pricing
The free plan is a real course, not a teaser: every lesson and every question, plus a graded sample of every format. Pro opens the whole problem bank.
Learn everything and try every format.
$0forever
No card needed.
Create a free accountThe full problem bank, every format.
$9/month
Billed $108 a year. Or $15 billed monthly.
Start your free trialQuestions
Straight answers to what people ask first. Still unsure? Start on the free plan; it doesn't expire.
Every lesson, staff-level deep dive and quick revise page, every multiple-choice question, the easy coding and SQL problems, and one graded sample of each premium format: a design project, a debugging task, a code review, a test-writing task and a system design problem. Mastery tracking, weak areas and spaced review work on the free plan too. It never expires.
Your first Pro subscription starts with a 7-day free trial. You choose monthly or yearly at checkout, and nothing is charged until the trial ends. Cancel before then and you pay nothing.
Yes. Open Profile, then Plan & billing, and cancel in a couple of clicks. You keep Pro until the end of the period you've paid for, and your progress stays on the free plan. You can also switch between monthly and yearly or download invoices there.
Yes. Pro is priced in US dollars, euros and British pounds. EU and UK prices include VAT; US sales tax is added at checkout where it applies. Payments are handled securely by Stripe.
Problem sites are great for algorithm practice. AI Proof Engineers adds what they don't track: a picture of your mastery per skill and concept, the misconception behind each wrong answer, spaced review, and formats beyond puzzles, such as design projects, code review, debugging, SQL tuning and system design. A chatbot can explain an idea, but it can't reliably tell you whether your solution is correct. AI Proof Engineers never asks one to: code runs against hidden tests, SQL is compared row by row, and designs are checked by rules.
You build the design from typed components (load balancers, caches, queues, databases and so on) with arrows marked sync or async, instance counts, capacity estimates and notes. The checks are rules that follow from the brief: traffic enters through a load balancer, there's no single point of failure, reads survive a cache outage, slow work goes through a queue. You see every check with what was wrong and why it matters, plus a reference design.
Java 21 or Python, your choice, and you can switch anytime. Coding problems, design projects, debugging tasks and code reviews come in both. SQL runs on PostgreSQL 18, and lessons show Java and Python side by side.
No. The courses go well past interview trivia: staff-level deep dives, concurrency, language versions and production trade-offs. It's built for engineers who want to know where their foundations are thin, whether that's for an interview next month or the job they already have.
Lessons, questions and review work in a phone browser. Coding, SQL and the design canvas are best on a laptop, where the editor has room. Native Android and iOS apps are on the way.
Submitted code runs in a short-lived, locked-down container, separate from the application and its database. You can download everything tied to your account as JSON, or delete the account, from Profile, then Data & privacy.
Your next step
Learn the idea, solve against real tests, and let the evidence tell you what to work on next. Start free today; upgrade when you want the full problem bank.
Free plan, no card · 7-day Pro trial · Cancel anytime