Pure AI Engineering Track

"The best AI engineers in 2025 are software engineers who know how to build with LLMs, not ML researchers who learned to code."

Core Principle: Build, Don't Train

Traditional ML Engineer

Data → Train Model → Deploy → Maintain

Applied AI Engineer (This Track)

API → Prompt → RAG/Agent → Deploy

What You'll Master

Prompt Engineering
Getting LLMs to do what you want
RAG Architecture
Connecting LLMs to knowledge
AI Agents
Building autonomous workflows
LLM Evaluation
Measuring quality at scale
LLMOps
Production AI systems

What You Won't Need

PyTorch / TensorFlow
Model training
Backpropagation math
GPU cluster management
Research papers

Target Companies

Type	Examples	Why Pure AI Works
Product Companies	Notion, Figma, Slack	Adding AI features
SaaS + AI	Intercom, Zendesk	AI-powered products
AI Startups	Dust, Jasper, Copy.ai	LLM applications
Enterprise	Salesforce, SAP	AI integrations
Consulting	McKinsey, BCG	AI solutions

6-Week Intensive Curriculum

1

Foundations of Building with LLMs

Master LLM APIs and prompt engineering at expert level

Day 1: Environment + First API Calls (4 hours)

# Create project
mkdir ai-engineer-portfolio && cd ai-engineer-portfolio
python -m venv venv && source venv/bin/activate

# Install essentials
pip install openai anthropic langchain langchain-openai
pip install python-dotenv pydantic

Day 2-3: Prompt Engineering Mastery (12 hours)

Role prompting
Few-shot examples
Chain of Thought
Output format control
Self-consistency patterns

Day 4-5: Structured Output + Function Calling (10 hours)

from pydantic import BaseModel
from openai import OpenAI

class EmailAnalysis(BaseModel):
    sentiment: str  # positive, negative, neutral
    urgency: str    # high, medium, low
    category: str
    key_points: list[str]

client = OpenAI()
response = client.beta.chat.completions.parse(
    model="gpt-4o-2024-08-06",
    messages=[...],
    response_format=EmailAnalysis
)

Day 6-7: Building a Complete Chatbot (8 hours)

Production-ready chatbot with conversation memory, streaming responses, and error handling.

Week 1 Deliverables

[ ] LLM APIs working (OpenAI + Anthropic)
[ ] 20+ prompt patterns practiced
[ ] Structured output with Pydantic
[ ] Function calling implemented
[ ] Complete chatbot with streaming

2

RAG Systems (Production-Ready)

Build RAG that actually works in production

Day 1-2: Vector Databases + Embeddings (8 hours)

Chroma (local)

import chromadb
client = chromadb.PersistentClient(path="./db")
collection = client.create_collection("docs")

Pinecone (production)

from pinecone import Pinecone
pc = Pinecone(api_key="...")
index = pc.Index("documents")

Day 3-4: Complete RAG Pipeline (10 hours)

Document loading (PDF, TXT, Web)
Chunking strategies
Embedding + storage
Query + retrieval chain

Day 5-6: Advanced RAG Techniques (10 hours)

Hybrid Search

BM25 + Vector retrieval

Reranking

Cohere Rerank, cross-encoders

Query Transformation

HyDE, multi-query

MMR

Maximum Marginal Relevance

Week 2 Deliverables

[ ] Vector database setup (Chroma + Pinecone)
[ ] Production RAG pipeline
[ ] Hybrid search + reranking
[ ] Streamlit RAG app with citations

3

AI Agents That Work

Build reliable agents for real-world tasks

Day 1-2: LangChain Agents (8 hours)

Build agents with custom tools for web search, calculations, database queries.

Day 3-4: LangGraph for Complex Workflows (10 hours)

from langgraph.graph import StateGraph, END

class AgentState(TypedDict):
    messages: Annotated[list, operator.add]
    current_task: str
    research_results: list

workflow = StateGraph(AgentState)
workflow.add_node("research", research_node)
workflow.add_node("write", write_node)
workflow.add_node("review", review_node)
app = workflow.compile()

Day 5-6: Multi-Agent Systems (8 hours)

CrewAI teams with specialized roles: Researcher, Writer, Critic agents working together.

Week 3 Deliverables

[ ] Agent with 4+ tools
[ ] LangGraph workflow with state
[ ] Multi-agent CrewAI system
[ ] Blog post: "Building Reliable AI Agents"

4

Evaluation + Production

Ship quality AI with confidence

Day 1-2: RAG Evaluation (8 hours)

from ragas import evaluate
from ragas.metrics import faithfulness, answer_relevancy, context_precision

result = evaluate(dataset, metrics=[
    faithfulness,
    answer_relevancy,
    context_precision,
    context_recall
])

Day 3-4: LLMOps + Monitoring (8 hours)

LangSmith integration
Custom metrics tracking
Cost monitoring

Day 5-6: Production Deployment (8 hours)

FastAPI server, streaming responses, Docker container, CI/CD with evals.

Week 4 Deliverables

[ ] RAGAS evaluation pipeline
[ ] LangSmith monitoring
[ ] Production FastAPI server
[ ] Docker deployment

5

Portfolio + Leadership Topics

Build impressive portfolio, prepare for Lead role

Day 1-3: Capstone Project (15 hours)

Option A

Enterprise RAG Platform

Option B

AI-Powered Support System

Option C

Research Assistant

Day 4-5: AI Leadership Topics (8 hours)

When to Use AI (Product Thinking)
AI Team Structure
Build vs Buy vs API
Ethics & Safety
Stakeholder Communication

Week 5 Deliverables

[ ] Capstone complete
[ ] 3 portfolio projects
[ ] Demo videos
[ ] Blog posts

6

Job Search Execution

Apply strategically, ace interviews

Application Materials

LEAD AI ENGINEER

SUMMARY
Senior engineer transitioning from iOS to AI. Built production LLM
applications including RAG systems and AI agents. Led teams of X engineers.

PROJECTS
- Production RAG System - 94% faithfulness score, hybrid search
- Multi-Agent Research Assistant - Reduced research time by 70%
- LLM Evaluation Pipeline - Integrated with CI/CD

Interview Prep Topics

System Design

"Design RAG for customer support"
"Design multi-agent document processing"

Technical

Implement RAG pipeline
Write prompt for complex task

Week 6 Deliverables

[ ] 30+ applications
[ ] Resume optimized
[ ] Interview prep complete
[ ] Mock interviews done

Tech Stack

# Core
pip install langchain langchain-openai langchain-anthropic
pip install llama-index chromadb pinecone-client

# Agents
pip install langgraph crewai

# Eval
pip install ragas deepeval

# Production
pip install fastapi uvicorn streamlit
pip install langsmith

Key Concepts Cheat Sheet

RAG

LLM + your documents

Embeddings

Text → numbers for search

Vector DB

Database for embeddings

Agent

LLM that uses tools

LangGraph

State machine for agents

RAGAS

Metrics for RAG quality

LangSmith

Monitoring for LLM apps

Success Metrics

By Week 6, You Should:

3 production-quality projects

Can build RAG in 2 hours

Can design agent systems

Understand evaluation deeply

Ready for Lead AI interviews