Are these based on real engineering roles?
Yes. The catalog is built around real engineering role patterns so the practice round feels closer to a live interview.
Company practice
Cognition mock interviews
Pick a role, answer follow-up questions out loud, and get a scored verdict after the interview.
4 rolesPythonAgentsAppSecAWS
Create custom interviewRoles to practice
Cognition
Infrastructure
Research Engineer, Infrastructure
Build the training and evaluation infrastructure behind Cognition's models for Devin, owning the data, compute, and harness that let researchers iterate quickly on post-training and agent capabilities. You'll bridge ML research and large-scale systems engineering. A technical interview would probe distributed training pipelines, large-scale data and eval harness design, and how you'd build infrastructure that makes model and agent experiments fast, reproducible, and cheap to run.
PythonPyTorchCUDACognition
Security
Security Engineer
Secure Cognition's products and infrastructure as Devin autonomously reads, writes, and executes code inside customer environments, owning the threat model for an AI software engineer with broad access. You'll harden sandboxed execution, cloud infrastructure, and the agent's permission boundaries. A technical interview would probe how you'd isolate and contain arbitrary agent-run code, design least-privilege access for an autonomous agent, and reason about the unique attack surface created by an AI that can modify and deploy software.
Cloud SecurityAppSecSandboxingCognition
DevOps / SRE
Site Reliability Engineer
Own the production reliability of Devin and Cognition's user-facing products along with the platform engineering that lets teams ship fast, covering SLOs, incident response, and on-call on one side and CI/CD, deployment infrastructure, and developer tooling on the other. You'll work with cloud infrastructure, Kubernetes, and infrastructure-as-code while writing real code. A technical interview would probe incident response and observability instincts, CI/CD and deployment design for fast-moving teams, and how you'd systematically reduce toil and define SLOs from a user's perspective.
KubernetesTerraformAWSCognition
ML / AI
Software Engineer
Design and ship the systems that power Devin's long-horizon task execution: tool use, context management, multi-step planning, subagent orchestration, and sandboxed code-execution environments. This is applied-AI systems work on getting an agent to reason reliably across thousands of lines of code, not feature plumbing. A technical interview would probe agent architecture and tool-use design, strategies for managing context over long-running tasks, and how you'd make multi-step agent behavior reliable and recoverable when individual steps fail.
PythonLLMsAgents
ExoForm is not affiliated with Cognition. This is an independent practice page.
FAQ
Is the interview voice-based?
Yes. ExoForm runs a live voice interview, asks follow-ups, and produces structured feedback after the session.
Can I try it for free?
Yes. You can start with the free interview allowance before upgrading for more practice.