k8s/cloud based llm container challenge

[commjoen]

Context

have a very small model hosted in a llama.cpp container (SmolLM2-135M-Instruct (Q4 GGUF is ~80-100MB, fits 256MB with room to spare), SmolLM2-360M-Instruct if you want a bit more coherence; that's more like a 512MB fit.) which hides a secret as part of the instruction

• For what category would the challenge be? (e.g. Docker, K8s, binary)
  AI, but with k8s infra

Note: the challenge needs to be implemented in minikub k8s, GCP, AWS and Azure

[commjoen]

will create container in #2689 with separate docker container and k8s setup