CrackML by @ml.with.umang
Interview questions / ML System Design
ML System Design interview question

Design a Bot-Post Detection Pipeline

Design an ML system to detect automated or coordinated bot-generated posts at platform scale.

hardsystem designEvidence 41/1001 source reportMeta

The 60-second answer

Define the abuse taxonomy and operational action separately from the score: downrank, challenge, review, or enforcement have different error costs. Combine content, account, behavioral, graph, device, and temporal signals; build labels from reviewed cases while guarding against adversarial label contamination.

Build the answer in this order

1
Frame the problem

Define the abuse taxonomy and operational action separately from the score: downrank, challenge, review, or enforcement have different error costs.

2
Design the data path

Combine content, account, behavioral, graph, device, and temporal signals; build labels from reviewed cases while guarding against adversarial label contamination.

3
Choose the modeling stack

Use layered detection with cheap rules/features, learned models, confidence thresholds, and human escalation for high-impact or uncertain cases.

4
Serve, evaluate, iterate

Monitor precision/recall by abuse slice, prevalence, reviewer disagreement, evasion patterns, latency, and enforcement reversals.

A useful interview mental model

This is the shape of a strong answer—not a script to memorize.

01Requirements
02Data
03Model / Retrieval
04Serving
05Monitor

Senior-level signal

  • Senior answers design for adaptive adversaries, policy versioning, and delayed/biased ground truth.
  • Discuss graph-level coordinated behavior and safe rollback if a model suddenly changes enforcement rates.

What the interviewer is really testing

Product framing, data design, modeling choices, serving constraints, reliability, evaluation, and explicit trade-offs.

Likely follow-up questions

What changes at 10× traffic or data volume?
Which failure mode would you monitor first in production?
How would you evaluate this offline and online before rollout?

Common weak-answer patterns

  • Jumping to a model before defining the product contract.
  • Listing components without bottlenecks, metrics, or failure handling.
  • Ignoring data quality, serving latency, monitoring, and iteration.