Enterprise-class RAG in C# and .NET - AI-powered document search
An implementation program covering the delivery of two complete enterprise-grade RAG applications and an AI agent for automating business processes centered on searching and analyzing large document collections. Participants master the full document processing pipeline - from embeddings and chunking, through semantic and hybrid search, up to scaling with ANN/HNSW/IVF/FAISS. The training combines intensive C# and .NET workshops with the practical implementation of enterprise-class extensions (rate limiting, guardrails, BIAS analysis, regression testing). The outcome is a production-ready RAG architecture and a foundation for further automation of document-centric processes.
Why this training?
Solutions based on LLM and RAG (Retrieval Augmented Generation) have become the implementation standard for organizations managing large document collections - regulations, contracts, reports, manuals. Simply “connecting ChatGPT” is not an enterprise-grade implementation. Real production systems require a precise document processing pipeline, a scalable search architecture, quality control mechanisms and integration with the organization’s existing technology stack.
This training guides participants through the full cycle of building two complete enterprise-grade RAG applications in C# and .NET - from the first line of code to production-ready extensions. It does not stop at a demo - it shows how to solve real performance problems with millions of documents, how to ensure guardrails and rate limiting, how to perform BIAS analysis and model regression tests.
What sets the program apart?
- Two complete RAG applications built from scratch during the training - not fragments or demos
- .NET/C# stack with Semantic Kernel as a first-class framework for orchestrating AI agents
- Scaling to millions of documents using ANN/HNSW/IVF and the FAISS library (100-1000x speedup)
- Hybrid search combining FTS (Full-Text Search) with semantic retrieval
- Enterprise extensions: rate limiting, guardrails, BIAS analysis, outlier detection, regression tests
- AI agent for SQL as a Semantic Kernel plugin - a foundation for further automation of business processes
Outcomes for the organization
After the training, the team is ready to independently deliver an enterprise-grade RAG system inside the organization - without dependency on external vendors. An architecture aligned with the existing .NET stack minimizes technical debt, and the ready-made control mechanisms (guardrails, rate limiting, regression tests) allow moving past the pilot stage and entering production with a system that meets the requirements of regulated industries.
Benefits
- Build two complete enterprise-grade RAG applications in a production-ready architecture
- Master the full document processing pipeline (chunking, embeddings, indexing, search)
- Deliver hybrid solutions combining full-text and semantic search
- Apply Semantic Kernel (.NET) to orchestrate AI agents and SQL plugins
- Design scalable RAG systems using ANN, HNSW, IVF and the FAISS library
- Implement enterprise-class extensions: rate limiting, guardrails, BIAS analysis, regression tests
- Establish a technology baseline for further automation of document-driven business processes
Who is this training for?
Prerequisites
- Hands-on knowledge of C# and the .NET platform
- Basic knowledge of relational databases and T-SQL
- General awareness of machine learning and large language models (LLM)
- Minimum 2 years of experience in commercial software development
Training program
Technology stack overview
- Workshop environment - core
- Alternative enterprise technologies
- Summary and rationale for the technology stack
Environment installation
- Hardware requirements
- Software installation and required components
- Graphical schema of the model's operation
Case: document search system in a bank
- Introduction to the system
- Business scenario
- Scope of documents
- The document search problem
- Solution architecture
- Data model
Database, DBO, FTS setup and document overview
- Database and DBO creation
- FTS (Full-Text Search)
- PDF document overview
- Preparing DBO for documents
- Target search process
- Loading test documents
Embeddings, vectors, methods
- Introduction
- Representing text as a vector
- Vector space
- Vector similarity
- Cosine similarity method
- Mathematical derivation
- Example cosine similarity calculation
- L2 normalization
- ANN - Approximate Nearest Neighbor
- Manual cosine similarity calculation
- Cosine similarity implementation in T-SQL
- Cosine similarity implementation in .NET C# (CLR)
- Case: performance test - SQL vs .NET C#
Pipeline construction (process)
- Introduction to the pipeline (process)
- Pipeline architecture (process)
- Purpose of splitting documents into fragments
- What is a chunk
- Sliding window
- Overlap
- Chunk granularity
- Segmentation strategy for legal regulations
- Chunking implementation in .NET C#
- Embedding models and their justification
- Embeddings implementation in .NET C#
Complete RAG project in C#
- Introduction to the approach
- RAG application implementation in C#
Document cleaning and updating
- Introduction
- Procedure implementation
- Connecting the procedure to the application
Case: semantic search (hybrid)
- Introduction to search methods
- Pipeline architecture (process) of the RAG system
- llama3 model
- RAG (hybrid) pipeline implementation
- Case: required documents list (PL)
Case: using Semantic Kernel (.NET)
- Introduction
- Semantic Kernel (.NET) and orchestration
- Target RAG implementation and orchestration
- Summary
- AI agent for SQL (plugin) v1
- AI agent for SQL (plugin) v2
The scaling problem in RAG systems
- Introduction
- Brute-force similarity = O(n)
- Bottlenecks
- Why 10k works and 1m does not
- Case: time measurement
- Case: how much RAM the embeddings database consumes
- Case: real-world test in a C# project
ANN, HNSW, IVF, FAISS: 100-1000x speedup
- ANN (Approximate Nearest Neighbor)
- HNSW (Hierarchical Navigable Small World)
- IVF (Inverted File Index)
- HNSW vs IVF
- FAISS library vs Vector DB
FAISS installation
- Introduction
- Python installation
- venv virtual environment
- FAISS installation
- FAISS installation verification
- Summary
RAG: enterprise level - implementation
- Final schema of the enterprise-class solution
- FAISS API construction (Flask)
- Integration with C# (our project)
- Testing
- Summary
Enterprise-class extensions in a RAG system
- Rate limiting
- Guardrails
- BIAS analysis
- Outlier detection
- Model regression tests
- Module summary
Delivery Methods
Online
- Convenience of participating from anywhere
- Interactive live sessions with trainer
- Materials available for 30 days
- No travel costs
On-site
- Direct contact with trainer and group
- Intensive hands-on workshops
- Networking with other participants
- Full focus on learning
Frequently asked questions
What are the prerequisites for this training?
We recommend hands-on knowledge of C# and the .NET platform, basic knowledge of relational databases and T-SQL, general awareness of ML/LLM concepts and at least 2 years of commercial software development experience.
What is the format and duration of this training?
The training lasts 4 days (32 training hours) and is available in online and on-site formats. Sessions run from 9:00 AM to 4:00 PM with breaks. We can customize the schedule to fit your team's needs.
Who is this training designed for?
This training is aimed at .NET/C# developers, system architects, digital transformation specialists and AI/ML engineers delivering RAG solutions in organizations across finance, insurance, legal, telecom, BPO and HoReCa sectors.
Will I receive a certificate?
Yes - every participant receives an EITT completion certificate together with a full set of training materials.
What technologies are used during the training?
The main stack is C#/.NET with Semantic Kernel, SQL Server (T-SQL), Python (venv) and the FAISS library with a Flask API. Throughout the training, two complete enterprise-grade RAG applications and an AI agent for SQL are built.
Request a quote
Funding Options
Check funding options for your company
Development Services Database
Up to 80% funding for SMEs from EU funds
Check availabilityNational Training Fund
Up to 100% funding for employers
Learn moreTrusted by
We train teams at Poland's largest companies
Interested in this training?
Contact us - we'll prepare an offer tailored to your organization's needs.