The ML Engineer's Path #3intermediate

Meta AI & Open Source ML

Meta has open-sourced more AI than any company: Llama, FAISS, SAM, Detectron2. Learn to run LLMs locally, fine-tune with LoRA, build vector search, and deploy RAG pipelines — all with zero API costs. Plus Microsoft ONNX, Hugging Face, and MLflow.

What You Build

A complete AI application using only open source: Llama + FAISS RAG pipeline with web frontend

Sections

Lessons

Estimated

Concepts

Prerequisites

PyTorch from Zero to Production

llamametafaissopen-sourceragfine-tuninghugging-faceonnx

Dev Jordan

walkthrough · casual

First 4 lessons are free (29% of the course). No login required.

The Open Source AI Revolution

Why Meta and the open source community are changing AI forever.

Meta's Open Source AI Strategy

20 min Build in Public

Tweet: Tweet: 'Meta released more open-source AI than anyone. Llama, FAISS, Detectron2, Segment Anything. Let's build with all of them.'

The Open Source ML Landscape

20 min

Llama: Your Own AI Model

Run, fine-tune, quantize, and serve Llama models.

Run Llama Locally

35 min Ship

Ship: Llama running on your machine, answering questions

Fine-Tune Llama with LoRA

40 min Ship Build in Public

Ship: Fine-tuned Llama model specialized for your domain

Tweet: Tweet: 'Just fine-tuned Llama on my own data using LoRA. Custom AI model, running locally, zero API cost.'

Preview — content blurs halfway through

Llama.cpp & GGUF: Quantized Inference

30 min

vLLM & Text Generation Inference

30 min

Unlock the rest of this course

$5 for full course or $25/mo for everything

See pricing

Meta's AI Toolbox

FAISS, SAM, Detectron2, and the Hugging Face ecosystem.

FAISS: Billion-Scale Vector Search

35 min

Segment Anything (SAM): Universal Segmentation

30 min

Detectron2: Object Detection & Segmentation

35 min

Hugging Face: The Model Hub

30 min

Open Source MLOps & RAG

Build production RAG pipelines with zero API costs.

Microsoft ONNX Runtime & Azure ML

30 min

Building RAG Pipelines with Open Source

40 min

Open Source MLOps: MLflow, Weights & Biases

30 min

Final Ship

Build a complete AI app using only open source tools.

Ship an Open Source AI App

50 min

Next Course

ML Engineering for Principal Engineers

You don't need to train models — you need to design the systems that train, serve, monitor, and scale them. GPU infrastructure, cost optimization, model serving patterns, ML CI/CD, security, team topology, and incident response. The system design interview for ML, but practical.

Meta AI & Open Source ML

What You Build

A complete AI application using only open source: Llama + FAISS RAG pipeline with web frontend