| Generative AI Internship Outline (9 Weeks) | ||||
| Course Title: Generative AI Duration: 2 Months (9 Weeks / 45 Working Days) Hours: 4 hours per day Total Contact Hours: 180 hours Total Trainees: 10 Instructor Name: Ayesha Azam |
||||
| Week | Day | Topic Details | Deliverables | |
| Week 1 – NLP Preprocessing | Day 1 | Intro to NLP, text structure, real-world applications | Notes/slides on NLP foundations | |
| Day 2 | Tokenization, stop words, stemming, lemmatization | Code snippets + examples | ||
| Day 3 | POS tagging, Named Entity Recognition (NER) | Annotated text example using spaCy/NLTK | ||
| Day 4 | Build complete preprocessing pipeline | Jupyter notebook: end-to-end pipeline | ||
| Day 5 | Apply pipeline to synthetic dataset | Processed dataset + Blog Post | ||
| Week 2 – Sequence Models & Transformers | Day 6 | RNNs, LSTMs, GRUs – architectures, vanishing gradients, temporal dependencies | Diagrams & notes on RNN/LSTM/GRU architectures | |
| Day 7 | Transformer architecture – attention, self-attention, positional encoding | Annotated Transformer diagram & concept map | ||
| Day 8 | Implement LSTM for text generation or classification | Training notebook + sample outputs | ||
| Day 9 | Fine-tune pretrained transformer (e.g., DistilBERT) on classification task | Fine-tuning notebook + model evaluation results | ||
| Day 10 | Compare RNN vs Transformer models – performance, training dynamics | Comparative report + PPT | ||
| Week 3: LLMs & Prompt Engineering | Day 11 | GPT, T5, BERT – comparison and evolution of LLMs | Comparison table + slides | |
| Day 12 | Tokenization schemes and embeddings | Demo notebook | ||
| Day 13 | Prompt engineering techniques | Prompt design exercise | ||
| Day 14 | Use OpenAI/HuggingFace APIs to build chatbot | Functional chatbot | ||
| Day 15 | Share chatbot and write Prompt Engineering deck | Working bot + PPT | ||
| Week 4: RLHF, Ethics, and Evaluation | Day 16 | Language model scaling & RLHF | Reading summary | |
| Day 17 | Ethical concerns: bias, hallucination, safety | Short blog post/reflection | ||
| Day 18 | Compare outputs of 2–3 LLMs | Comparative table + PPTs | ||
| Day 19 | Evaluate outputs with BLEU, ROUGE, perplexity | Metric score report | ||
| Day 20 | Finalize and submit evaluation write-up | Evaluation Report | ||
| Week 5: Literature Review and Project Ideation | Day 21 | Conducting technical literature reviews | Guide + paper selection | |
| Day 22 | Summarize key LLM and Gen AI papers | 3–5 paper summaries | ||
| Day 23 | Present and discuss selected papers | Group presentation deck | ||
| Day 24 | Identify gaps and brainstorm ideas | Idea sheet: 2–3 project ideas per team | ||
| Day 25 | Finalize mini-project topics | Mini-project proposal | ||
| Week 6: Retrieval-Augmented Generation (RAG) | Day 26 | Introduction to RAG: architecture & purpose | RAG report | |
| Day 27 | Tools: LangChain, LlamaIndex, Haystack | Tool comparison report | ||
| Day 28 | Implement basic RAG pipeline | Q&A pipeline notebook | ||
| Day 29 | Evaluate and improve the RAG system | Evaluation metrics + scores | ||
| Day 30 | Submit RAG system demo | Working RAG demo | ||
| Week 7: Agentic AI | Day 31 | Introduction to Agentic AI: autonomy, planning | Blog post or summary note | |
| Day 32 | Tooling: LangChain agents, AutoGPT | Functional demo: simple agent | ||
| Day 33 | Agent memory and planning | Enhanced agent code | ||
| Day 34 | Use agents for project-specific workflows | Customized agent for mini-project | ||
| Day 35 | Share agent + document experience | Agent demo + project note | ||
| Week 8: LLM Trainings | Day 36 | Introduction to fine-tuning: full, adapter-based, PEFT methods | Slide deck or notes on fine-tuning methods | |
| Day 37 | LoRA & QLoRA: parameter-efficient fine-tuning concepts | Annotated diagrams + configuration explanation | ||
| Day 38 | Implement LoRA/QLoRA fine-tuning on small LLM | Fine-tuning notebook | ||
| Day 39 | Evaluate performance: memory usage, speed, output quality | Evaluation results + comparison chart | ||
| Day 40 | Document your fine-tuning workflow for reproducibility | Write-up/report + final model artifact | ||
| Week 9: Final Project Presentations, Documentation, & Reflection | Day 41 | Technical documentation best practices: reproducibility, versioning, testing | Documentation checklist + project README | |
| Day 42 | Final polishing: code cleanup, test cases, runtime validation | Final codebase + testing logs | ||
| Day 43 | Prepare final presentations (technical + impact + demo) | Slide deck and demo video | ||
| Day 44 | Final project presentations (individuals or teams) | Live or recorded presentation | ||
| Day 45 | Reflection, feedback session, blog wrap-up, and closing | feedback form submission | ||