To those who are left behind in the AI rush

Source: Dev.to

Cheat Sheet: Key AI Concepts

• NLP (Natural Language Processing) – AI field that enables computers to understand, interpret, and generate human language.
• LLM (Large Language Model) – Trained on massive text corpora to predict the next word, enabling chat, writing, and summarization.
• LMM (Large Multimodal Model) – Extends LLMs to handle images, audio, video, and other modalities.
• Foundational Model – A massive, general‑purpose model (e.g., GPT‑4) that serves as a base for many downstream tasks.
• Multimodal – Ability to process and connect different data types (e.g., describing an image in text).
• Prompt Engineering – Crafting specific instructions to elicit the best responses from an AI model.
• Finetune – Further training a pre‑trained model on a smaller, task‑specific dataset.
• RAG (Retrieval‑Augmented Generation) – Connecting an AI to an external knowledge base so it can fetch real‑time facts, reducing hallucinations.
• Embedding Model – Converts text or images into numeric vectors that capture meaning, enabling similarity comparisons.
• AI Engineering – Building complete applications using ready‑made AI models, focusing on integration, security, cost, and usability.
• ML Engineering (Machine Learning Engineering) – Designing, training, and optimizing models from scratch, including data cleaning and algorithmic improvements.
• Agentic AI – AI that can act autonomously: breaking goals into steps, using tools (e.g., web search, booking), and completing tasks.
• MLOps (Machine Learning Operations) – Practices that keep AI models updated, monitored, and reliable in production.

Real‑World Project Pipeline (Customer Support Agent)

AI Engineer Stack: Common Tools

• Orchestration (workflow glue) – LangChain, LlamaIndex
• Vector Databases (RAG storage) – Pinecone, Weaviate, Chroma
• App Builders & UI – Streamlit, Gradio
• Evaluation & Observability – LangSmith, Arize Phoenix
• Local Development – Ollama, LM Studio

Leading Tools & Models Overview

Foundational Models

• Text & General Reasoning: OpenAI GPT‑4o, Anthropic Claude 3.5 Sonnet, Google Gemini 1.5 Pro, Meta Llama 3 (open‑source leader).
• Image Generation: Stable Diffusion, Midjourney
• Video Generation: Synthesia (talking heads), Google Veo (cinematic clips)

AI Applications & Assistants

• Coding Assistants – GitHub Copilot
• Meeting Automation – Fireflies.ai (record, transcribe, summarize)
• Search & Research – Perplexity AI (conversational search with citations)
• Workflow Assistants – Lindy (multi‑step business workflows)

RAG Tools

• Frameworks – LlamaIndex (indexing & retrieval), LangChain (application logic)
• Vector Databases – Pinecone (managed), Weaviate (open‑source), Chroma
• Evaluation – LangSmith, Arize Phoenix

Agentic AI

• Multi‑Agent Teams – CrewAI, Microsoft AutoGen (role‑specific AI agents)
• Complex Workflows – LangGraph (task control, human‑approval steps)
• Autonomous Coding – Goose, Claude Code (edit files, run tests)

Other Important Engineering Tools

• Local AI Runners – Ollama (run open‑source models like Llama 3 locally)
• Model Optimization – DSPy (automatic prompt optimization)