1. Target and capacity Role Applied Scientist Machine Learning Engineer Research Scientist Research Engineer Target level L4 / mid-level L5 / senior L6 / staff L7 / principal Senior principal / distinguished Primary domain General / not sure LLM and agent systems Recommendations, search, and ranking ML platform, training, and serving Research and modeling Post-training, RL, and agent environments Alignment, safety, and model behavior Multimodal, vision, speech, or robotics General product ML Hours available each week About 5 hours About 8 hours About 12 hours Weeks until the loop 2 weeks 4 weeks 6 weeks 8 weeks 12+ weeks
2. Actual interview rounds Select the formats the recruiter described, not every topic you might study. Domains such as LLMs and recommendations belong above, not here.
Use typical rounds for this role ML breadth Mechanisms, model choices, metrics, and “why” follow-ups. ML implementation / coding ML primitives, evaluation systems, debugging, and resource-aware implementation under time pressure. Agentic codebase Navigate an unfamiliar ML codebase, direct an AI coding agent, review its work, and ship a tested change. ML system design Requirements, data, modeling, evaluation, serving, monitoring, and iteration. Project deep-dive Ownership, technical decisions, failures, evidence, influence, and reflection. Technical strategy Shared boundaries, investment order, delegated authority, decision checkpoints, reversibility, and retained technical depth. Technical project presentation Present one consequential ML project, then defend its decisions, evidence, failures, and impact under interruption. Behavioral / leadership Conflict, prioritization, failure, influence, mentoring, and collaboration. Values & mission Reason through a real ethical tension, explain a changed belief, and show principled disagreement without slogans. Product & experimentation Metrics, experiment design, causal threats, guardrails, and ship decisions. Research depth / paper critique Hypotheses, ablations, evidence quality, derivations, and research judgment. Research work sample Turn a black-box observation into hypotheses, discriminating probes, measured evidence, and a short readout. Math & statistics oral Rapid derivations, assumptions, sanity checks, interpretation, and changed-assumption follow-ups. Systems / infrastructure Training and serving scale, reliability, bottlenecks, and cost trade-offs.
Rounds outside this site's curriculum Select only formats confirmed for your loop. These affect readiness, but mlmentorship intentionally does not duplicate dedicated coding, SQL, or general systems resources.
General algorithms / DSA Use a dedicated coding resource and record a recent timed baseline. Not in my loop No recent timed baseline Workable timed baseline Practical software, OOP, or concurrency Practice a progressive stateful task outside this ML-specific library. Not in my loop No recent timed baseline Workable timed baseline SQL and data manipulation Use a dedicated SQL resource with realistic event and experiment tables. Not in my loop No recent timed baseline Workable timed baseline General distributed-system design Prepare storage, queues, consistency, partitioning, and failure semantics separately. Not in my loop No recent timed baseline Workable timed baseline