Large Language Models (LLMs)
LLMs are foundation models trained on massive text corpora that can understand and generate human-like text. They power chatbots, code assistants, content generation and much more.
- GPT-4 / GPT-4o (OpenAI) - Multimodal model for text, code and image understanding
- Claude (Anthropic) - Safety-focused LLM with strong reasoning and long context windows
- Gemini (Google) - Multimodal AI model integrated across Google products
- LLaMA & Open-Source Models - Meta's open-weight models enabling local and custom deployments
- Mistral & Mixtral - Efficient open-source models with strong performance per parameter
RAG (Retrieval-Augmented Generation)
RAG combines the power of LLMs with external knowledge retrieval, allowing models to answer questions grounded in your own data without hallucination.
- Vector Databases - Pinecone, Weaviate, Chroma, Qdrant for storing and searching embeddings
- Embedding Models - OpenAI Embeddings, Sentence Transformers, Cohere Embed for text vectorization
- Chunking Strategies - Fixed-size, semantic, recursive and document-aware splitting
- Hybrid Search - Combining dense vector search with sparse keyword search (BM25) for better recall
- Re-Ranking - Cross-encoder models for improving retrieval precision
Prompt Engineering
Prompt engineering is the art and science of crafting effective instructions for LLMs to produce desired outputs consistently.
- Zero-Shot & Few-Shot - Providing examples to guide model behavior without fine-tuning
- Chain-of-Thought (CoT) - Encouraging step-by-step reasoning for complex problems
- System Prompts - Setting persona, constraints and output format for consistent results
- Tool Use / Function Calling - Enabling LLMs to interact with external APIs and databases
- Structured Output - Constraining LLM output to JSON, XML or other formats
LLM Application Frameworks
Frameworks and tools for building production-ready applications powered by LLMs.
- LangChain - Framework for building LLM-powered chains, agents and retrieval systems
- LlamaIndex - Data framework for connecting LLMs with structured and unstructured data
- Semantic Kernel - Microsoft's SDK for integrating LLMs into applications
- Streamlit & Gradio - Rapid prototyping frameworks for AI-powered web interfaces
- Guardrails - Validation and safety frameworks for LLM outputs (NeMo Guardrails, Guardrails AI)
Fine-Tuning & Model Customization
Fine-tuning adapts pre-trained models to specific domains or tasks, improving performance on specialized use cases.
- LoRA & QLoRA - Parameter-efficient fine-tuning for reducing compute and memory costs
- RLHF (Reinforcement Learning from Human Feedback) - Aligning models with human preferences
- Instruction Tuning - Training models to follow specific instructions and formats
- Domain Adaptation - Specializing models for healthcare, finance, legal or technical domains
- Evaluation - BLEU, ROUGE, human evaluation and LLM-as-judge for quality assessment
AI Agents & Autonomous Systems
- ReAct Pattern - Reasoning and acting in an interleaved loop for complex task completion
- Multi-Agent Systems - Multiple specialized agents collaborating on complex workflows
- Tool Use - Agents that can browse the web, execute code, query databases and call APIs
- Memory Systems - Short-term and long-term memory for maintaining context across interactions
- Planning & Decomposition - Breaking complex goals into executable sub-tasks