free · no sign-up · every answer worked out
Free exercises with worked answers
Each exercise is a real situation you will meet at work: a question, a hint if you want one, and a full answer that shows every step of the sum. The numbers come from model files, datasheets and documentation we read on 13 November 2026, and from our own GPU and Wazuh labs, and every page lists its sources.
LLM inference exercises
Sizing GPU memory, counting users, pricing a million tokens and splitting a big model, worked with real model files and our own GPU runs. They come from the course Hands-On Inference Engineering.
Which LLMs can I run on a Mac with 36, 48 or 128 GB of memory?
How much of a Mac's memory the GPU may use, real file sizes, and the KV cache: which local models fit in 36, 48 and 128 GB.
Which LLMs fit on an RTX 4090 with 24 GB of VRAM?
What fits in 24 GB with llama.cpp and vLLM, checked against our own RTX 4090 runs and vLLM's start-up log.
How many users can one L40S or H100 serve with Llama 3.1 8B or 70B?
Memory sets how many conversations fit at once; speed sets how many feel quick. Worked for Llama 3.1 8B and 70B on L40S and H100.
What does 1 million tokens cost to self-host on RunPod compared with an API?
Our measured RTX 4090 speed, dated RunPod and API prices, and the one number that decides it: how busy the GPU is.
Tensor parallelism vs pipeline parallelism for a 70B model: which one, and why?
Llama 3.1 70B on two H100s and on four L40S: what each split puts on each GPU, what they send each other, and which to pick.
SOC analyst exercises
Triaging alerts, writing a Wazuh rule and mapping logs to MITRE ATT&CK, using alerts and logs from our own Wazuh cloud lab. They come from the course AI SOC Analyst.
Alert triage exercise: five real Wazuh alerts, which are an attack and which are noise?
Five alerts from our own cloud lab: one typo, one new account and one attack that got in. Decide each, then see the worked answer.
Write a Wazuh custom rule that alerts when a new user account can log in
A small rule that sits under Wazuh's rule 5902 and raises level 12 when a new account can log in, tested line by line with wazuh-logtest 4.14.10.
MITRE ATT&CK mapping exercise: map these real log lines to techniques
Real Linux and Windows lines from our Wazuh lab, mapped to ATT&CK techniques and compared with what Wazuh itself says.
Want to change the numbers yourself? The free tools include the GPU memory and cost calculators these exercises use.