32dots HEIDELBERG AI
Curriculum › Local & Private Models › Ollama course
EXTENDED COURSE

Ollama

Ollama runs open AI models locally from the command line and exposes them through an OpenAI-compatible API on port 11434 — a private endpoint your scripts, apps, or whole lab can call without sending data to the cloud or paying per token. Downloading and running models on your own hardware is free and open source (optional paid cloud tiers add larger hosted models). In five short lessons you go from install to a shared, always-on private LLM endpoint, with biology-flavoured examples throughout.

1Lessons5step-by-step, ~90 min each

Setup

  1. 00 Install Ollama & run your first model

Basics

  1. 01 Pull, manage & compare models from the CLI
  2. 02 Call the OpenAI-compatible local API

Advanced

  1. 03 Power Hermes with your local model
  2. 04 Serve your whole lab from one headless machine
2Cheat sheetcopy-ready expressions
3Examples4what people built

Internal tools & ops2

Ollama AI-native SME Scientist
Course starter

Offline PII redaction before cloud upload

A short Python script loops over HR or patient records, sends each row to a locally running Llama 3 model via the Ollama REST API, and asks it to replace names, addresses, and ID numbers with placeholders — the cleaned file is written locally before anything goes to an external service.

Sensitive records are de-identified on the machine, satisfying data-governance requirements without involving a third-party API.

Try it yourself

Run `ollama pull llama3`, then POST each row to `localhost:11434/api/generate` with prompt: "Replace all PII in the following text with [REDACTED_<TYPE>]. Return only the cleaned text."

Ollama Founder AI-native SME
Course starter

Local code review assistant wired into the editor

A VS Code extension (or simple shell alias) pipes a selected code block to a locally running CodeLlama model via `ollama run`, asking for a brief review covering bugs, edge cases, and readability — results appear in the terminal or a side panel.

A free, always-available code reviewer that works on proprietary code offline, with no API subscription needed.

Try it yourself

Run `ollama pull codellama`, then in a terminal: `cat myfile.py | ollama run codellama "Review this code for bugs and suggest improvements."` Wire to a keybinding for repeated use.

Forms, surveys & feedback1

Ollama Small biz Founder
Course starter

Batch-classify customer feedback without an API bill

A CSV of hundreds of survey responses is fed row-by-row to a local Mistral model through the Ollama API; each response gets a category label (positive / negative / feature-request / bug) written back to a new column — no cloud API key, no token costs.

Thousands of rows classified overnight on a laptop for free, with no data ever leaving the machine.

Try it yourself

Run `ollama pull mistral`, then iterate the CSV in Python and POST each `response_text` to `localhost:11434/api/generate` asking for a single category label; write labels back to the DataFrame.

Knowledge & docs1

Ollama AI-native SME Founder
In the gallery

Summarise confidential board reports on a plane

Board packs, legal memos, or NDA-protected documents are pasted into a local prompt; a Llama 3 model running in Ollama returns executive summaries and key action items — the laptop needs no internet connection at any point.

Decision-relevant summaries are ready before landing, with zero data leakage risk and no dependency on airport Wi-Fi.

Try it yourself

Run `ollama pull llama3`, then `ollama run llama3` and paste the document text with prompt: "Summarise in 5 bullet points, then list action items with owners."