AI/ML Engineering
The AI Software Engineer role is applied, not academic. This module is about wiring models into products: tokens, embeddings, retrieval, agents, serving trade-offs, and the evaluation loops that keep quality from silently regressing.
Start the first lesson →The AI Engineer's Mental Model
Where the AI SWE sits between research and product, the shape of an LLM call, and the vocabulary the rest of the module builds on.
Not started2Tokens, Sampling & Prompting
How text becomes tokens, what temperature and top-p actually do, and prompt structures that make outputs reliable.
Not started3Embeddings & Vector Search
How meaning becomes geometry. Drag a query around a 2D vector space and watch cosine similarity rank the neighbors.
Not started4Vector Databases & ANN Indexes
Exact search does not scale. HNSW, IVF, and the recall-vs-latency knobs behind every production vector store.
Not started5Retrieval-Augmented Generation
Ground a model in your own data. Trace a question through chunking, retrieval, and generation stage by stage.
Not started6Prompting vs. RAG vs. Fine-tuning
The three levers for changing model behavior, what each is good and bad at, and how to choose without guessing.
Not started7Agents, Tools & Function Calling
Let a model take actions. The reason–act loop, tool schemas, and the failure modes that make agents hard in production.
Not started8Evaluation & Guardrails
You cannot improve what you cannot measure. Offline evals, LLM-as-judge, online signals, and safety guardrails.
Not started9Serving, Latency & Cost
Batching, KV-caches, streaming, and quantization. The engineering that decides whether your feature is affordable.
Not started10MLOps & the Model Lifecycle
Versioning data, models, and prompts; safe rollouts; and catching drift before users do. Keeping a live system healthy.
Not started