# ML Inference Lab for OpenScience

> OpenScience inference serve+bench: OpenAI-compat serve, p50/p99, throughput. BYOK only.

- Agent: OpenScience
- Plan: Pro (2 vCPU, 6 GiB RAM, 50 GiB disk)
- Forks: 0 · Stars: 0
- Creator: https://www.jurniti.com/creator-ca0ff2
- Tags: inference, latency, ml, openscience, serving, throughput, vllm
- Requires config (you provide): ANTHROPIC_API_KEY, GEMINI_API_KEY, HF_TOKEN, OPENAI_API_KEY, OPENROUTER_API_KEY
- Skills: ml-inference-bench, ml-inference-serve

**To fork:** open https://www.jurniti.com/templates/ml-inference-lab-for-openscience and use the Fork action — it starts a Stripe Checkout for a new VM seeded with this template's persist disk (same flat plan price, no fork fee).

## README

# OpenScience ML Inference Lab

Fork → BYOK → pick a model + serve shape (online OpenAI-compatible API vs batch).

## What you get
- **ml** agent default for inference work
- Skills: `ml-inference-serve` + `ml-inference-bench`
- Layout: `serve/`, `bench/`, `clients/`, `runbooks/` under `~/ml-inference-lab/`

## BYOK
```bash
export ANTHROPIC_API_KEY=sk-ant-…
export HF_TOKEN=hf_…  # optional
```

## Demo
Ask the **ml** agent to design an OpenAI-compatible serve + p50/p99 bench scaffold.

Click **Open OpenScience** on your VM for the browser workspace.