Forward Deployed Engineer (FDE) Master Program: GenAI, RAG, Agents & Enterprise AI – Live Training

Forward Deployed Engineer (FDE) Master Program: GenAI, RAG, Agents & Enterprise AI – Live Training

Prerequisites:

Section 1 — FDE Foundations & Mindset

• What is a Forward Deployed Engineer?
• FDE vs Software Engineer, Data Engineer, ML Engineer, Solutions Architect
• Origins of the role — Palantir, Databricks, OpenAI, Anthropic, Scale AI
• The four technical pillars: Data, AI, Software, Cloud & Platform
• The four core skills: Critical Thinking, Problem Solving, Decision Making, Communication
• The forces in orbit — scoping, ambiguity, cost, customer environment, scalability, governance
• Capability x Context — why context is the exponent
• Customer obsession, ownership mentality and bias for action
• Comfort with ambiguity and first-principles thinking
• Common mistakes FDEs make in the first 90 days

Section 2 — Critical Thinking

• Critical thinking fundamentals — the 8 traits (Analytical, Objective, Logical, Skeptical, Inquisitive, Open-minded, Reflective, Decisive)
• Critical thinking rules — separating symptom from root cause
• RED Critical Thinking Framework (Recognise assumptions, Evaluate arguments, Draw conclusions)
• Fishbone / Ishikawa root cause analysis
• 5 Whys and causal chain analysis
• Managing assumptions and assumption registers
• Asking the right questions — discovery questioning technique
• Risk identification and pre-mortem analysis
• Cognitive biases in engineering decisions
• Eisenhower Matrix and prioritisation under pressure
• Trade-off reasoning with incomplete information
• Decision making — reversible vs irreversible decisions

Section 3 — Problem Solving & Structured Thinking

• Converting vague business problems into technical requirements
• Scoping under uncertainty — bounding the problem before coding
• Estimation techniques for ambiguous work
• SCAMPER for solution ideation
• IDEA framework for structured problem solving
• Rapid prototyping — 0 to working demo
• Delivering iterative value — thin vertical slices
• Prioritisation under constrained time and budget
• Production incident response and debugging under pressure
• Unblocking yourself in a hostile environment (no docs, no access, no data)

Section 4 — Communication & Stakeholder Management

• Communicating with non-technical stakeholders
• Problem translation — business pain to technical spec and back
• Four Gates of thoughtful communication (True, Kind, Necessary Now, Makes a Difference)
• PREP framework for structured verbal responses
• Running discovery workshops and technical interviews
• Technical storytelling for executives
• Rapid demo culture — show, do not tell
• Communicating trade-offs and saying no well
• Building trust with stakeholders on a live engagement
• Written communication — status reports, decision logs, handover docs
• Cross-team collaboration and managing up

Section 5 — Generative AI Foundations

• What is Generative AI? AI vs ML vs Deep Learning vs GenAI
• Discriminative vs generative models
• Neural networks and transformer architecture (practitioner view)
• Attention mechanisms — self-attention and multi-head attention
• Positional encoding
• Tokens and tokenisation
• Context window, memory and session management
• Pre-training vs fine-tuning vs RAG
• Open vs closed models — when to use what
• Popular LLMs — Claude, GPT, Gemini, LLaMA, Mistral
• Reasoning models and extended thinking
• Model scaling laws and capability trends
• Hallucination — why it happens and first-line defenses
• Cost, latency and quality trade-offs
• Lab: Build your first AI app (API call + prompt + parsed response)

Section 6 — Prompt & Context Engineering

• Prompt engineering — the FDE superpower
• Markdown and XML structured prompt frameworks
• Zero-shot, one-shot and few-shot prompting
• Chain-of-thought, prompt chaining and multi-turn design
• System prompts and role design
• Structured output — JSON mode and schema enforcement
• Model controls — temperature, top-k, top-p, max tokens, stop sequences
• Enterprise prompt patterns — classification, extraction, summarisation, routing
• Prompt evaluation and iteration methodology
• Prompt guardrails and injection defenses
• Context engineering vs prompt engineering
• The context window as a budget — anatomy of context
• Short-term vs long-term memory; retrieval as context
• Compaction, summarisation and tool-result curation
• Failure modes — lost-in-the-middle and context rot
• Mini project: Enterprise prompt library

Section 7 — Embeddings & Vector Databases

• What are embeddings? Dimensionality and vector space
• Embedding models — OpenAI, Cohere, Sentence Transformers
• Similarity metrics — cosine, dot product, Euclidean
• Chunking strategies for enterprise documents
• Chunk overlap and context preservation
• Indexing methods — flat, IVF, HNSW
• Vector databases — Pinecone, FAISS, Weaviate, pgvector
• Vector store operations — CRUD, upserts and re-indexing
• Retrieval performance and cost optimisation
• Lab: Index a document set and run semantic search

Section 8 — Retrieval-Augmented Generation (RAG)

• What is RAG and why FDEs rely on it
• RAG architecture — ingestion, retrieval, generation
• Data ingestion pipelines — PDFs, databases, APIs, SharePoint
• Embedding pipeline design and incremental refresh
• Retrieval strategies — top-k, MMR, re-ranking
• Metadata filtering and contextual retrieval
• Advanced RAG — self-query, HyDE, corrective RAG, multi-hop, GraphRAG
• RAG vs fine-tuning — decision criteria
• RAG evaluation — faithfulness, answer relevance, context precision and recall
• Citations and source attribution
• Mini project: Enterprise knowledge assistant with RAG

Section 9 — AI Agents & Tool Use

• What is an AI agent? Perception, reasoning, action
• Agent architecture — planner, memory, tools
• Tool calling and function execution
• ReAct pattern — reason plus act loop
• Agentic design patterns — planning, reflection, reflexion
• AI workflows vs agentic AI — when each fits
• Single-agent vs multi-agent trade-offs
• Multi-agent architecture — supervisor-worker and peer patterns
• Agent frameworks — LangChain, LangGraph, CrewAI
• Agent observability, tracing and debugging
• Lab: Research agent with tool calling

Section 10 — Model Context Protocol (MCP) & Integrations

• What is MCP and why it matters for enterprise integration
• MCP architecture — hosts, clients and servers
• MCP primitives — resources, tools, prompts, sampling
• Transport layers — stdio vs SSE / HTTP
• MCP vs function calling vs REST API
• Build / invoke an MCP server for enterprise tools
• Real-world MCP integrations — filesystem, Git, Slack, browser, databases
• AI workflow and automation patterns for enterprise operations

Section 11 — Fine-Tuning & Model Customization

• When to fine-tune vs prompt vs RAG
• Supervised fine-tuning and instruction tuning
• LoRA and PEFT techniques
• Dataset preparation, splits and leakage control
• Model evaluation metrics for customised models
• Cost, safety and bias considerations in customisation

Section 12 — GenAI Evaluation, Observability & Guardrails

• AI observability — what to monitor and why
• Tracing, logging and metrics for AI applications
• Evaluation frameworks — LLM-as-judge and RAGAS
• Golden datasets, regression suites and CI for prompts
• Guardrails — input and output validation, safety filters
• Prompt injection, jailbreaks and AI security threats
• AI red teaming and adversarial testing
• Hallucination mitigation strategies in production
• Human-in-the-loop system design
• Lab: Add observability and guardrails to a RAG pipeline

Section 13 — GenAI Governance, Risk & Responsible AI

• AI safety fundamentals and risk taxonomy for GenAI
• Bias, fairness and model evaluation for harm
• Data privacy, PII handling and compliance (GDPR, DPDP, HIPAA)
• Enterprise AI governance frameworks and approval gates
• AI regulation and compliance landscape (EU AI Act and beyond)
• Responsible AI in client engagements — practical guardrails from day one