Forward Deployed Engineer (FDE) & Generative AI Systems

Master Curriculum & Engineering Reference for Building Production-Grade Enterprise AI Systems.

This repository contains structured lecture notes, system architectures, mathematical foundations, paper deep-dives, and production blueprints for Forward Deployed Engineers (FDEs) and Generative AI Systems Architects.

---

🎯 Repository Overview

• Role & Mindset: Moving from client-requested "chatbots" (proposed solutions) to true enterprise business problem discovery and resolution.
• Mathematical & Theoretical Depth: Step-by-step proofs (e.g., Attention variance scaling factor $\frac{1}{\sqrt{d_k}}$, sinusoidal positional relative offsets, $O(1)$ path length complexity).
• Paper Deep-Dives: Comprehensive breakdown of seminal research papers including "Attention Is All You Need" (Vaswani et al., 2017).
• Production Architecture: Design criteria for Grounded, Observable, Secure, and Actionable enterprise AI systems.
• Hands-on Implementation: Production microservices built with Node.js, Express, TypeScript, and modern AI SDKs.

---

📚 Curriculum Navigation

• 📖 Main Repository Index & Curriculum Roadmap
• 📂 Lecture 01 Index & Notes | 📄 Complete Document
• 📂 Lecture 02 Index & Notes | 📄 Complete Document
• 📂 Lecture 03 Index & Notes | 📄 Complete Document | 💻 Codebase | 🛠️ Implementation Guide
• 📂 Lecture 04 Index & Notes | 📄 Complete Document | 💻 Codebase | 🛠️ Implementation Guide

---

📖 Lecture Highlights

Lecture 01: Introduction to FDE & Generative AI Foundations

E-commerce case study (50k daily queries), Problem vs Solution discovery, Symbolic AI & Expert Systems, Machine Learning paradigm shifts, Statistical N-grams, RNN hidden state decay, Transformer Attention mechanism, and Enterprise Production AI Stack topology.

Lecture 02: How LLMs Work (From Tokens to Transformers)

Autoregressive next-token prediction, Subword tokenization (BPE/WordPiece), Dense Vector Embeddings $\mathbb{R}^{d_{\text{model}}}$, Positional Encodings, Scaled Dot-Product Self-Attention ($Q, K, V$), Transformer multi-block state propagation, Logits projection head, Softmax, Temperature scaling, and autoregressive streaming.

Lecture 03: Talking to an LLM from Your Application

ChatGPT vs Raw LLM (Car vs Engine analogy), Deterministic tools (Calculator, DB, Weather), Knowledge cutoff limits, The 4 Invariant Request Components (Where, Who, Which Model, What), Token economics (Prefill vs Decode, cost & latency), Node.js + Express + TypeScript microservice setup, Zod validation, OpenAI client singletons, response telemetry audit, and prompt injection defense via multi-role architecture.

Lecture 04: Building a Chatbot: Context, Memory, Streaming & Hallucinations

Statelessness of HTTP LLM endpoints, the illusion of conversational continuity, coreference and pronoun binding failure, message roles (system, user, assistant), prompt contamination anti-patterns, the 4 Pillars of system prompts, context window budget constraints, quadratic token accumulation problem ($O(N^2)$ bloat), lost-in-the-middle phenomenon, sliding window and rolling summarization in TypeScript, Parametric Knowledge ($\Theta$) vs In-Context Working Memory ($X_{\text{prompt}}$), autoregressive token streaming via Server-Sent Events (SSE), fluency vs truth orthogonality, coherent fabrication vs benign nonsense, model sycophancy (Aditya Tandon sorting law experiment), and the 7-layer defense-in-depth architecture.