Questions reported from Cognition machine learning engineer interviews. This bank tracks 4 questions asked at Cognition (3 with free model answers). Practice them live with DeepOffer's AI voice interviewer.
An agent loop alternates model planning, typed tool calls, observation, and stopping.
Start with workload, users, data, quality metrics, latency, throughput, freshness, reliability, and cost targets.
Loop over model output, validate a typed tool request, execute through a bounded dispatcher, append the observation, and stop on completion, error, or budget.
Get asked follow-ups live, then receive a scored report — like a real MLE interview loop.
Start AI mock interview