IT · Generative AI
Generative AI and AI Application Development Internship
Design useful LLM-powered applications with retrieval, evaluation and safety built in.
A software-focused generative AI pathway: learners build a grounded AI application and learn to evaluate quality, protect data and handle failure modes.
Program overview
Eligible branches: Computer Science and Engineering, Information Technology, Software-related engineering, Other branches with basic programming and API familiarity
Target learner level: Learners with basic Python or JavaScript experience who want to build model-powered software.
Prerequisites: Basic programming experience in Python or JavaScript · Comfort with APIs and JSON is helpful
Technologies and tools
Large language model concepts · prompt design · model APIs · JSON schemas · embeddings · vector search · retrieval-augmented generation · evaluation and safety
Practical assignments
- Create structured model requests with validation and failure handling
- Prepare documents and implement retrieval with source citations
- Build a small evaluation set and test relevance, safety and regressions
Capstone project
Build a source-grounded generative AI application with evaluation, data safeguards and clear human fallback paths.
Expected deliverables
- Working application and architecture notes
- Document preprocessing and retrieval flow
- Evaluation set and results summary
- Safety and privacy checklist
- Demonstration and limitations report
Learning outcomes
- Build an application using model APIs and structured outputs
- Implement retrieval-augmented generation with source grounding
- Evaluate quality, safety, latency and cost trade-offs
Week-by-week curriculum
The program follows a 13-week core pathway. Weeks 14–15 are optional advanced work when the project scope calls for it.
Weeks 1–2
Orientation, fundamentals and setup
- Orientation, AI-use boundaries, API keys and secure environment setup
- LLM concepts, tokens, context windows, prompting and structured output basics
Weeks 3–5
Core concepts and guided learning
- Prompt design, system instructions, templates and function/tool calling
- Model APIs, streaming responses, JSON schemas and error handling
- Embeddings, vector search and document preprocessing
Weeks 6–8
Intermediate concepts and practical exercises
- Build retrieval-augmented generation with citations and chunking
- Conversation state, tool workflows and application UX
- Create a small evaluation set; test relevance, faithfulness and regressions
Weeks 9–11
Advanced topics, specialist tools and project development
- Guardrails, prompt-injection defense, privacy and sensitive-data handling
- Model selection, caching, rate limits, latency and cost optimization
- Develop a domain-grounded AI capstone with human fallback paths
Weeks 12–13
Capstone completion, testing, documentation and evaluation
- Complete the application, source-grounded responses and evaluation report
- Test edge cases, safety failures, access boundaries and deployment behavior
- Document architecture, model choices, limitations and present the demo
Weeks 14–15 · Optional
Advanced extension and final presentation
- Add an agentic tool workflow or multimodal feature with clear controls
- Expand evaluation coverage and deliver an improved final presentation
Frequently asked questions
Is prior experience required?
The prerequisites listed on this page describe helpful preparation. The opening weeks include a guided refresher; learners should be ready to practise consistently and complete project work.
Are model API charges included?
No model-provider account or paid API usage is promised. Any provider access and associated cost must be confirmed before use.