LLM Applications

This directory covers the full lifecycle of Large Language Model (LLM) applications, from the underlying transformer mechanics to advanced autonomous agents.

Learning Path

1. LLM Mechanics

The "How it works" section.

• Transformer Architecture (Self-Attention, Positional Encodings).

• Scaling Laws (Chinchilla optimality).

• Inference Optimization (KV-Caching).

• Tokenization & Decoding strategies.

2. RAG: Retrieval-Augmented Generation

Connecting LLMs to external data.

• Indexing & Vector Databases (HNSW).

• Advanced Retrieval (Re-ranking, HyDE, Multi-query).

• GraphRAG & Knowledge Graphs.

• Evaluation via RAGAS Triad.

3. Tuning & Optimization

Adapting models for specific tasks.

• SFT (Supervised Fine-Tuning).

• PEFT (Parameter-Efficient Fine-Tuning) using LoRA/QLoRA.

• Alignment (RLHF vs DPO).

• Model Quantization (4-bit, bitsandbytes).

4. Agentic Workflows

Autonomous LLM agents.

• Reasoning Patterns (CoT, ReAct).

• Tool Use & Function Calling.

• Multi-Agent Orchestration.

• Compound AI Systems.

---

Interview Readiness

Every file in this directory includes a dedicated Interview Questions section covering both theoretical "Why" and practical "How" for production-grade LLM engineering.