went through the Salesforce MLE loop for the Einstein AI/ML platform team. i'll be honest about what was good and what was annoying.
the loop:
5 rounds. recruiter screen, then:
coding - standard algorithm problem in CoderPad. medium difficulty, graph traversal. if you're an MLE with engineering chops this should be fine. they're not trying to compete with the FAANG grinding, but they do want to see you can code fluently.
ML system design - this was the most interesting round by far. they gave me: design a model to predict which CRM opportunities are likely to close in the next 30 days. this is essentially the Einstein Sales Cloud feature, which made sense. they want you to think through: problem framing, feature engineering (what signals from a CRM make sense), training data construction (labels from historical close data, survivorship bias, leakage), model choice + trade-offs (gradient boosting vs. something more interpretable given sales reps need to trust the output), evaluation metrics (precision/recall at a threshold vs. ranking), and deployment/monitoring.
the deployment/monitoring part caught a few people off guard based on my debriefing conversation. know how you'd detect model degradation in production and what you'd do about it.
ML depth - a senior MLE asked deep questions about one of my past projects. they wanted to go 3 levels deep: what did you do, why did you make that choice, what would you do differently now. bring a project you genuinely understand end to end. do not hand-wave.
behavioral - STAR stories. specifically: a time you shipped something that was technically imperfect but right for the situation, a time you disagreed with your manager on a modeling approach, a time you had to work across teams to get something deployed.
hiring manager - relaxed. team context, growth path, how they think about ML maturity in the org.
what i'd prep: ML system design is the differentiator at Salesforce for MLE. they're thinking about real B2B CRM use cases, so it helps to think about prediction tasks in that domain: lead scoring, churn, next best action. feature engineering from structured CRM data (activity logs, deal stage transitions, email patterns). and seriously prep your deployment/monitoring story.
comp was competitive with senior MLE at comparable enterprise companies. my offer was $185k base + RSUs, Bay Area/remote hybrid.