Maximum Programmer-Problem Matching

Each problem has a list of tags, and each programmer has a list of skills. A programmer is compatible with a problem when the programmer's skills and the problem's tags share at least one identical string.
Match programmers to problems so that each programmer and each problem appears in at most one pair. Return the maximum possible number of compatible pairs.

Function
maximumProblemMatches(problemTags: List<List<String>>, programmerSkills: List<List<String>>) → int

Examples
Example 1
problemTags = [["java"], ["python", "sql"], ["go"]]
programmerSkills = [["java", "python"], ["java"], ["rust", "go"]]
return = 3
The first programmer can take the Python problem, the second programmer can take the Java problem, and the third programmer can take the Go problem. These three pairs use every programmer and problem once, so no larger matching is possible.
Example 2
problemTags = [["java"], ["python"], ["sql"]]
programmerSkills = [["java", "python"], ["python"]]
return = 2
Match the first programmer to the Java problem and the second programmer to the Python problem. Only two programmers are available, so the maximum is 2.

Build a bipartite graph. Programmers on the left, problems on the right, with an edge whenever their string sets overlap. The answer is the maximum matching. Standard approach: map each tag to the list of problems that carry it, so you build edges without comparing every pair of lists. Then run Kuhn's algorithm. For each programmer, DFS to find a free problem, or try to reassign the problem's current owner somewhere else. Track a visited set per attempt. The classic pitfall is greedy first-fit, which can lock a flexible programmer onto a problem a stricter one needed. Another is forgetting to reset visited between programmers. Dedupe tags into sets first. Hopcroft-Karp is the upgrade if sizes are large. If the DFS recursion tangles during the live OA, StealthCoder is the hedge that hands you a working matching template.

If this hits your live OA and you blank, StealthCoder solves it in seconds, invisible to the proctor.