Generative AI Training in Anna Nagar

• Introduction to AI, Machine Learning, Deep Learning, and Generative AI.
• Understanding Transformer architecture and attention mechanisms.
• Introduction to popular LLMs such as GPT-4, Claude, Llama 3, and Mistral.
• Understanding tokens, tokenization, and text embeddings.
• AI ethics, bias, and responsible AI practices.
• Setting up Python, Jupyter Notebook, and Hugging Face.

Beginner Build: A Python-based customer support chatbot using an open-source LLM.

• Understanding Zero-shot, Few-shot, and Chain-of-Thought prompting.
• Creating reusable prompts with dynamic inputs.
• Generating structured outputs in JSON and XML formats.
• Testing and improving prompt quality.
• Reducing incorrect AI responses and hallucinations.
• Managing and tracking prompts using LangChain.

Beginner → Intermediate Build: A Smart Resume Parser that extracts candidate details from PDFs into structured JSON.

• Introduction to LangChain and LlamaIndex.
• Connecting multiple LLM tasks using sequential and router chains.
• Managing conversation history and AI memory.
• Connecting AI applications with external APIs.
• Building basic AI agents using the ReAct framework.
• Handling errors and retrying failed AI requests.

Intermediate Build: A Multi-Document Q&A Assistant that searches and answers questions from uploaded files.

• Understanding vector embeddings and similarity search.
• Introduction to Pinecone, Milvus, ChromaDB, and Qdrant.
• Preparing documents and splitting them into smaller chunks.
• Understanding keyword search and semantic search.
• Combining keyword and vector-based search.
• Improving search results using metadata filters.
• Measuring search accuracy using Hit Rate and MRR.

Intermediate Build: A ChromaDB-powered semantic search engine for a company knowledge base.

• Understanding basic RAG and advanced RAG architectures.
• Improving search results using query rewriting and multi-query retrieval.
• Understanding HyDE and advanced retrieval techniques.
• Re-ranking search results using Cross-Encoders and BGE-Reranker.
• Improving responses through contextual compression.
• Introduction to GraphRAG and knowledge graphs.
• Evaluating RAG applications using RAGAS and TruLens.

Intermediate → Advanced Build: An Enterprise Financial Analyst Bot using advanced RAG, re-ranking, and RAGAS evaluation.

• Introduction to multimodal AI models such as GPT-4V, LLaVA, and Stable Diffusion.
• Understanding Vision-Language Models (VLMs).
• Building text-to-image and image-to-text applications.
• Extracting speech and text from audio using Whisper.
• Understanding multimodal RAG for searching images and audio.
• Managing hardware requirements and processing speed.

Intermediate Build: A Visual Document Analyzer that extracts data, tables, and charts from scanned invoices.

• Understanding AI agents and the ReAct framework.
• Building multi-agent applications using CrewAI and AutoGen.
• Connecting AI agents with external tools and APIs.
• Implementing function calling and tool integration.
• Managing short-term and long-term agent memory.
• Implementing human approval and AI safety controls.
• Understanding Model Context Protocol (MCP).

Advanced Build: A Multi-Agent Research Assistant where specialized agents collaborate to research and report.

• Understanding when to use fine-tuning, RAG, or prompt engineering.
• Introduction to Supervised Fine-Tuning (SFT).
• Understanding LoRA and QLoRA for efficient model training.
• Preparing custom training datasets using JSONL and ChatML.
• Fine-tuning models using Hugging Face Transformers and TRL.
• Introduction to RLHF and Direct Preference Optimization (DPO).
• Testing and comparing fine-tuned models with original models.

Advanced Build: Fine-tune a Llama 3 model on a custom customer support dataset using QLoRA.

• Generative AI applications in healthcare and medical documentation.
• Using AI for legal document analysis and contract review.
• Applying AI in financial data analysis and fraud-related insights.
• Detecting and removing personal information (PII) from documents.
• Building secure, locally hosted AI applications.
• Understanding enterprise AI use cases and industry applications.

Advanced Build: A Legal Contract Reviewer that highlights risks, identifies clauses, and redacts sensitive data.

• Understanding the LLMOps lifecycle.
• Evaluating AI responses using BLEU, ROUGE, and BERTScore.
• Introduction to LLM-as-a-Judge evaluation.
• Tracking experiments using MLflow and Weights & Biases.
• Building automated testing pipelines for AI applications.
• Monitoring model performance and detecting data drift.
• Managing API costs, token usage, and response caching.

Advanced Build: An LLM Evaluation Dashboard using MLflow and RAGAS to track AI response quality.

• Building AI APIs using FastAPI and Pydantic.
• Packaging applications using Docker.
• Introduction to Kubernetes for AI application deployment.
• Deploying AI applications using AWS Lambda and Google Cloud Run.
• Improving model performance using vLLM and TensorRT-LLM.
• Understanding model quantization and inference optimization.
• Implementing Redis caching and load balancing.

Expert Build: Deploy a production-ready RAG API using FastAPI, Docker, and vLLM on the cloud.

• Selecting a real-world problem for the final AI project.
• Designing AI application architecture and selecting technologies.
• Using Git and GitHub for version control and team collaboration.
• Preparing project documentation and architecture diagrams.
• Building a professional GitHub portfolio and AI-focused resume.
• Practising technical interviews and AI system design questions.
• Understanding career opportunities in Generative AI and AI Engineering.

Expert Build: A fully deployed Enterprise AI Copilot (e.g., HR Assistant or Code Reviewer) with documentation and a demo.