AI Architect

AI Platform Architect

5 Free 45 Pro

Sign in for free to access all 5 free questions.

Showing 12 of 50 questions

ai-01 beginner

Free

AI Workstation Forge

Configure your OS/GPU setup, then build a reproducible local AI dev environment on a plain ₹40-50k laptop (no discrete GPU required), and ship your first self-built tool: a model benchmarking CLI.

Beginner 780m est.

ai-02 beginner

Free

Prompt Systems & Structured Output Contracts

Treat prompts as versioned, tested code: JSON-schema-constrained output and a prompt regression suite that runs in CI and fails the build on drift, not a chatbot demo.

Beginner 480m est.

ai-03 beginner

Free

Retrieval-Augmented Generation, v1

Build a minimal RAG service: chunking, embeddings, a real vector store, a retrieval API, and evaluate retrieval quality with recall@k, not by eyeballing a couple of answers.

Beginner 600m est.

ai-04 beginner

Free

Container & Ship an AI Service

Take the RAG service from Project 3 to production packaging standard: multi-stage Docker build, non-root user, health checks, and a full docker-compose stack.

Beginner 360m est.

ai-05 beginner

Free

Tool-Calling Agent, v1

A single-agent function-calling loop with 3-4 real tools, loop/cost guardrails, and full tool-call logging: an agent that cannot run away from you.

Beginner 480m est.

ai-06 beginner

Instrument the RAG and agent services from Projects 3-5 with OpenTelemetry traces, structured logs, and latency/token/cost metrics on a live dashboard. Treat LLM calls like any other production dependency.

Beginner 480m est.

ai-07 beginner

Evaluation Harness

An automated eval pipeline: golden dataset, LLM-as-judge plus deterministic checks, that gates a merge on quality regression, proven by a deliberately-broken PR it correctly blocks.

Beginner 480m est.

ai-08 beginner

CI/CD for an AI Service

A full pipeline for the Project 3-7 service: lint, test, eval-gate, build, push, deploy to staging, rollback. Proven with one real rollback drill, not just a green checkmark.

Beginner 480m est.

ai-09 intermediate

Self-Hosted Model Serving at Scale

Deploy vLLM on a rented cloud GPU, benchmark throughput and latency against Ollama on CPU, tune batching and concurrency, and expose an OpenAI-compatible API. The first of only 4 projects in the whole program that need a GPU.

Intermediate 420m est.

ai-10 intermediate

GPU Scheduling on Kubernetes

Stand up a Kubernetes cluster with the NVIDIA device plugin, GPU-aware scheduling and quotas, and deploy a model server with real probes and HPA. The second of 4 GPU-bound projects, and the one that structurally cannot run local-only.

Intermediate 480m est.

ai-11 intermediate

API Gateway & AuthN/Z for AI Endpoints

Put a real API gateway in front of the platform: API-key auth, per-tenant rate limits, and payload size limits. The boundary that turns a service into something other people can safely call.

Intermediate 420m est.

ai-12 intermediate

Multi-Model Routing & Fallback

A router choosing between local and hosted models on cost, latency, and availability, with automatic fallback and circuit breakers. Proven by surviving a simulated provider outage.

Intermediate 420m est.

AI Architect

AI Workstation Forge

Prompt Systems & Structured Output Contracts

Retrieval-Augmented Generation, v1

Container & Ship an AI Service

Tool-Calling Agent, v1

Observability for AI Services

Evaluation Harness

CI/CD for an AI Service

Self-Hosted Model Serving at Scale

GPU Scheduling on Kubernetes

API Gateway & AuthN/Z for AI Endpoints

Multi-Model Routing & Fallback