Anthropic Claude SWE Interview | CTC $350K | Please Help

You are working as a Software Engineer at Anthropic on an internal analytics system for Claude.

The Claude product team runs N feature launches across M engineering teams.
For every feature launch i, team j receives A[i][j] impact points based on how much that team contributed to the launch.

You are allowed to choose any subset of feature launches to include in a final internal impact report for Claude.

For the selected launches, each team’s total impact score is the sum of its impact points across all chosen launches.

Formally, if the selected subset of launches is S, then the total impact score of team j is:

score[j] = sum of A[i][j] for all i in S

If no launch is selected, every team has impact score 0.

Anthropic leadership wants to analyze how much separation can be created between Claude engineering teams in the final report.

For a team i, define its separation score as the sum of absolute differences between its total impact score and every other team’s total impact score:

separation(i) = |score[1] - score[i]| + |score[2] - score[i]| + ... + |score[M] - score[i]|

The final report’s separation value is the maximum separation score among all teams:

max separation(i) over all teams i

Your task is to choose a subset of Claude feature launches such that this final report separation value is maximized.

Return the maximum possible separation value.