Forward Deployed Engineering (FDE): Build Production-Ready AI Applications using Python, FastAPI, LLMs, RAG, AI Agents, Streamlit & Docker – Live Training

Module 1: FDE Mindset and Problem Solving (3 Hours)

• What Forward Deployed Engineers (FDEs) do in real-world scenarios
• Understanding business problems and converting them into technical solutions
• Breaking down complex requirements into manageable components
• Understanding constraints and trade-offs
• System thinking: Input → Processing → Output
• Designing practical AI application workflows

Module 2: Writing Production-Ready Python (3 Hours)

• Writing clean, modular Python functions
• Structuring Python projects
• Organizing files and separating logic
• Working with JSON data
• Reading and processing API responses
• Basic exception handling and debugging

Module 3: Backend Development with FastAPI (4 Hours)

• Understanding APIs and Request-Response architecture
• HTTP methods (GET, POST)
• Designing REST APIs using FastAPI
• Working with JSON input and output
• Adding business logic
• API testing using FastAPI Interactive Documentation
• Error handling and validation

Use Case

• Build a backend service with multiple API endpoints.

Module 4: Data Flow in Applications (2 Hours)

• Understanding how data moves through an application
• Using in-memory storage
• File-based storage using JSON
• Reading and writing application data
• Connecting backend APIs with stored data

Use Case

• Extend the backend application to store and retrieve information.

Module 5: AI & LLM Applications (8 Hours)

• Understanding how LLM APIs work
• Prompt structuring and prompt iteration
• Prompt engineering fundamentals
• Understanding variability in model outputs
• Temperature and randomness
• Hallucinations and limitations
• Practical evaluation techniques
• Choosing between Machine Learning and LLMs
• Cost, latency and reliability considerations

Use Cases

1. Email Routing Automation

Students will build an AI-powered email routing system capable of:

• Detecting customer intent
• Categorizing emails
• Routing requests automatically
• Creating structured outputs using LLMs

2. Multimodal AI Application

• Input:
• Product Image
• Output:
• Product Category

Students will learn:

• Image understanding using LLMs
• Prompt-based image classification
• Evaluating classification performance

Module 6: Retrieval-Augmented Generation (RAG) (4 Hours)

• Why LLMs require external knowledge
• RAG architecture and workflow
• PDF ingestion
• Text extraction
• Chunking strategies
• Fixed-size chunking
• Sentence-based chunking
• Recursive chunking
• Embeddings
• Cosine similarity
• Retrieval process
• Context construction
• Generation using retrieved context
• Common RAG failure cases
• Hallucination prevention

Use Case

• Build an end-to-end Document Question Answering System capable of answering questions using information from uploaded PDF documents.

Module 7: Building AI Agents (2 Hours)

• What is an AI Agent?
• Difference between an LLM and an AI Agent
• Understanding agent workflows
• Tool calling concepts
• Decision-making using LLMs
• Designing simple task-oriented AI Agents
• Limitations and best practices

Use Case

• Build a simple AI Research Assistant that can:
• Accept a user query
• Decide which tool or function to use
• Retrieve information
• Generate a final response based on tool outputs
• Students will understand how modern AI Agents combine reasoning with external tools to automate multi-step tasks.

Module 8: UI Integration with Streamlit (2 Hours)

• Building simple user interfaces
• Text input
• Buttons
• Displaying outputs
• Connecting Streamlit with backend APIs
• Connecting Streamlit with AI applications

Use Case

• Build a complete frontend for the AI application developed throughout the course.

Module 9: Packaging with Docker (2 Hours)

• Organizing project structure
• Understanding Docker
• Images vs Containers
• Writing a Dockerfile
• Building Docker images
• Running applications using Docker
• Packaging AI applications for deployment

Use Case

• Package the AI application into a Docker image and run it locally using Docker.

Capstone Project (2 Hours Guided + Additional Work Outside Sessions)

Students will work individually or in groups of 2–3 to design and build a complete end-to-end AI application.

Each project should include:

• Backend API (FastAPI)
• AI/LLM functionality
• Optional RAG or AI Agent capabilities where applicable
• Streamlit User Interface
• Docker packaging (Recommended)

Example Project Ideas:

• Intelligent Resume Analyzer
  Customer Support Assistant
  Policy Document Q&A System
  Contract Review Assistant
  Meeting Notes Summarizer
  AI Research Assistant
  Invoice Information Extraction System
  Legal Document Assistant
  Customer Feedback Analyzer
  Enterprise Knowledge Assistant

Students will:

• Define the problem statement
• Design the system architecture
• Build the complete solution
• Test and evaluate the application
• Present the final project with a live demonstration