Learn the AI stack without leaving your Kotlin brain behind.
You already know Kotlin, Android architecture, and how to ship. This series builds the AI product engineering stack — Python, LLM APIs, RAG, tool calling, agents, LangGraph, evaluation, and deployment — on top of that foundation, one topic at a time, with runnable code and Kotlin-to-Python comparisons throughout.
Sourced directly from real AI product engineering work.
| Track | Topics | Output |
|---|---|---|
| Foundations | Python bridge, LLM APIs, prompt/context engineering | You can call an LLM and control its output reliably. |
| Core patterns | Structured outputs, tool calling, embeddings, RAG, agents, LangGraph, evaluation | You can build a grounded, tool-using, evaluated AI feature. |
| Backend & deployment | FastAPI, pgvector, Docker/Cloud Run/Vertex AI | Your AI feature runs as a real, deployable service. |
| Capstone | End-to-end AI product, AI-assisted engineering workflows | A working decision-support product, and a faster dev loop. |
| Bonus | RAG on Android | A mobile client that uploads documents and streams cited answers. |
Foundations
Python for Kotlin Developers
Syntax, types, collections, classes, null handling, async, and packaging — mapped from what you already know.
LLM APIs Fundamentals
Gemini, OpenAI, and Claude: chat completions, streaming, tokens, and cost.
Prompt & Context Engineering
System prompts, few-shot examples, context windows, and failure modes.
Core LLM app patterns
Structured Outputs
JSON mode and schema-constrained generation.
Function / Tool Calling
Giving an LLM callable tools and closing the request/response loop.
Embeddings & Vector Search
What embeddings are, similarity search, and pgvector basics.
RAG
Chunking, retrieval, and grounding answers end-to-end.
Agentic Workflows
Agent loops, planning, memory, and multi-step execution.
LangGraph
Stateful graph-based orchestration for agents.
AI Evaluation
Evaluating LLM outputs and RAG pipelines systematically.