Generative AI and Agentic AI Course: Core Gen AI Engineering

Hands-onExplore a pretrained foundation model using Hugging Face. Perform summarization and compare model outputs. Inspect a prepared fine-tuning example and compare base-model versus adapted-model outputs.

Hands-onExtract structured information from unstructured text. Validate outputs against a JSON schema. Build a prompt test set and compare variants for accuracy, consistency, and handling of missing information.

Hands-onBuild an API-connected conversational assistant. Add streaming, conversation history, structured responses, and error handling. Implement a reusable application workflow using LangChain or LlamaIndex.

Hands-onGenerate embeddings for an enterprise document collection. Store and search embeddings using one selected vector store. Compare chunking strategies and metadata filters using a retrieval test set.

Hands-onBuild a company-policy knowledge bot that answers questions from uploaded PDFs. Add source citations and an insufficient-evidence response. Compare a baseline RAG pipeline with an enhanced retrieval pipeline and document the results.

Hands-onTest the RAG application for hallucinations, unsafe prompts, prompt injection, and sensitive-information exposure. Add input/output validation, redaction, refusal rules, and audit logs. Create a risk register and regression test suite.

Hands-onIdentify and scope an agent use case in a business workflow. Build a tool-using assistant that retrieves weather and destination information. Connect a prepared MCP tool and demonstrate a basic A2A exchange using supplied scaffolding. Add tool validation and approval for consequential actions.

Hands-onBuild a multi-step travel-planning or research assistant. Integrate APIs or mock services for external information. Add checkpointing, conditional routing, retry handling, and an approval step. Test recovery from a failed tool call.

Hands-onBuild a researcher–analyst–reviewer workflow using LangGraph. Track agent handoffs, tool calls, evidence, failures, and execution costs. Add task-completion checks and execution limits. Package the application behind an API and prepare deployment configuration.

Hands-onSelect one project: 1. AI-Powered Research Assistant: multi-agent research, summarization, and evidence-linked reporting. 2. Enterprise Knowledge Bot: policy-grounded RAG assistant with citations and access controls. 3. Customer Support Agent: FAQ retrieval, API integration, ticket routing, and human-approved actions. 4. No-Code AI Workflow Builder: n8n-based business automation with approval, logging, and error handling. Deliver a working prototype, evaluation report, risk register, deployment package, and demonstration.