Apply bagging, boosting, and stacking on the same dataset, compare metrics, and quantify ensemble lift over single models

Evaluate computational cost vs. performance gain for each ensemble technique and recommend deployment feasibility

Build a feed-forward neural network using Keras/PyTorch, achieve a specified validation loss, and document architecture choices.

Evaluate overfitting via learning-curve analysis and implement regularization (dropout/L2) to meet generalization targets.

Learners will apply LLMs to generate first-draft executive briefs that summarize model insights and refine prompts to meet specified ROUGE or BLEU scores.

Learners will create comprehensive data-to-text pipelines that combine SQL, Python, and LLM APIs to automatically transform KPI (Key Performance Indicator ) tables into narrative summaries.

Learners will fine-tune small LLMs on company FAQs and measure improvement in response relevance through systematic human evaluation.

Learners will evaluate cost vs. latency trade-offs between open-source and commercial LLMs for real-time chat applications through systematic analysis.

Build a chatbot prototype using RAG (retrieval-augmented generation) and measure user satisfaction through SUS survey.

Evaluate dialog-flow metrics (fallback rate, turn length) and iterate on intent-matching rules.

Apply named-entity recognition to extract key terms from support tickets and quantify precision/recall.

Evaluate two vectorization techniques (TF-IDF vs. embeddings) on a text-classification task.

You build an end-to-end AI-powered insights application that combines ensemble modeling with LLM-driven explanation generation. You train and evaluate an ensemble model to predict customer health scores, extract feature importance, and integrate a large language model to generate natural language explanations for each prediction. The final deliverable is a functioning application with a simple interface and technical documentation suitable for a development team.