System Design Studio
Pick a real interview problem. Draft the architecture on a drag-and-connect canvas. Submit and get an AI-graded review scored against the rubric — exactly like a staff loop.
Deep Research Agent
Design an agent that takes an open-ended query, plans sub-questions, searches and reads many web sources, and produces a faithfully cited synthesis.
Autonomous Coding Agent
Design an agent that takes a coding task, navigates a repo, edits code, runs tests, and opens a PR — safely.
Multi-Agent Orchestration
Design a system that coordinates multiple specialized agents to complete a complex task — and justify when multi-agent is worth it.
Computer-Use / Browser Agent
Design an agent that controls a browser/GUI to complete web tasks, where errors compound and the page content is untrusted.
Enterprise Knowledge Agent
Design a Q&A assistant over internal company knowledge (docs, wikis, tickets) with strict per-user permissions.
Long-Term Memory System
Give an agent persistent memory across sessions — deciding what to remember, keeping it consistent, and forgetting.
Real-Time Voice Agent
Design a low-latency speech-to-speech conversational agent with natural turn-taking.
Eval & Guardrail Platform
Design a platform to evaluate LLM/agent quality and enforce safety guardrails in production.
Conversational Analytics (Text-to-SQL)
Let non-technical users ask questions in plain English and get correct answers from a data warehouse with thousands of tables.
Customer-Support Triage & Resolution Agent
Design an agent that triages inbound support tickets, resolves what it can (including actions like refunds), and escalates the rest.
AI Coding Copilot (Inline Completion at Scale)
Design the inline gray-text code-completion system for millions of developers — instant, cheap, and high quality.
LLM Gateway / Model Router
Design an internal gateway every product team calls instead of hitting model providers directly.
Intelligent Document Processing & Workflow Automation
Design an agent that processes incoming documents (invoices/claims), extracts and validates data, and pushes it downstream — replacing a manual team.
Proactive Personal Assistant (Calendar + Email)
Design an assistant that manages a user's calendar and email — schedules meetings, drafts replies, and surfaces things proactively.
Trust & Safety / Content Moderation at Scale
Design LLM-based moderation for a platform with millions of posts/day — detect and act on harmful content under tight latency.
AI On-Call / Incident-Response (AIOps) Agent
Design an agent that responds to alerts: investigates, finds root cause, and proposes or executes remediation — without making things worse.
Synthetic Data Generation & Curation
Design a pipeline that generates synthetic training/eval data at scale — diverse, high-quality, and uncontaminated.
Conversational Commerce / Shopping Agent
Design a conversational shopping agent that understands needs, recommends real products, and helps complete a purchase.
- 30% structure. Rubric must-haves checked on your diagram. In Structured mode they tick off live as you draw.
- 70% AI review. A staff-level reviewer grades architecture, completeness, data flow, safety and scalability.
- Iterate. Every attempt is kept, so you can see your score move.