ISM 428: Project in Applied Artificial Intelligence and Information Systems
3 Credits | Prerequisite: ISM 422 – Deep Learning for Business
Course Description
ISM 428 is the capstone course for the Applied Artificial Intelligence and Information Systems program. Students synthesize knowledge acquired from ISM 250 (Intro to AI Literacy and Prompt Engineering), ISM 315 (Python Programming), ISM 324 (Database Management), ISM 325 (Systems Analysis & Design), ISM 326 (Data Provision for AI), ISM 360 (AI for Business), and ISM 422 (Deep Learning for Business).
Students complete one major applied AI project by selecting one of three pathways:
- Implementing a Large Language Model — build, train, or fine-tune a generative model (text, image, or video) using Learn Generative AI with PyTorch by Mark Liu.
- Applied AI Analytics Project — use AI/ML tools (traditional ML with Python, LLM APIs, text analytics, or generative modeling) to analyze a real-world dataset and produce actionable insight.
- AI-Driven Business Process Automation — design and implement an applied AI agent or Retrieval-Augmented Generation (RAG) pipeline to automate or re-engineer an existing business process.
Students work individually or in pairs. The course culminates in a formal written report, working prototype, and a professional presentation suitable for employers, competitions, or graduate school portfolios.
Required Text
Learn Generative AI with PyTorch, Mark Liu, Manning Publications, 2024
(Provided to students; selected chapters used depending on project pathway.)
Source: course-licensed PDF.
Course Learning Objectives
- Integrate knowledge from programming, database management, AI literacy, and deep learning into a real AI system.
- Build, train, fine-tune, or deploy a generative AI model for a practical purpose.
- Design AI-enabled business workflows using RAG pipelines, prompt engineering, embeddings, and automation tools.
- Collect, prepare, and document real-world data for machine learning tasks.
- Produce a professional AI report with reproducible code, documentation, and ethical assessment.
- Communicate technical findings to a business audience.
Course Format
The course is structured around project development. Each week includes:
- Short lecture or technical workshop
- Instructor or peer code review
- Studio time to develop the project
- Milestone submission
Assessment & Grade Distribution
- Project Proposal – 10%
- Milestone 1: Data & Architecture – 15%
- Milestone 2: Prototype Working System – 20%
- Milestone 3: Evaluation & Improvements – 15%
- Final Technical Report (15–25 pages) – 20%
- Final Presentation + Demo – 20%
Weekly Schedule (15 Weeks)
| Week | Topics | Deliverables |
|---|
| 1 |
Course overview, project pathways, example projects, ethical considerations in AI. |
Team formation; draft ideas. |
| 2 |
Proposal workshop; quick review of Python, PyTorch, LangChain, RAG workflows, embedding models. |
Project Proposal Due |
| 3 |
LLM Pathway: Chapters 1–3 (GANs) overview — from Mark Liu.
Analytics Pathway: dataset selection, problem framing.
RAG Pathway: vector databases, document loaders, chunking strategies. |
Dataset identified and approved. |
| 4 |
Project architecture design — model pipeline, data flows, evaluation plan. |
Milestone 1 Due: Data + Architecture |
| 5 |
LLM Pathway: Chapters 4–7 (Image GANs, VAEs).
Analytics: EDA, feature engineering, exploratory modeling.
RAG: embeddings, FAISS, Chroma, and LlamaIndex pipelines. |
Progress check. |
| 6–7 |
Development Sprint 1 — students build the earliest working prototype. |
|
| 8 |
Milestone 2 Workshop: debugging, performance improvements, GPU usage. |
Milestone 2 Due: Prototype |
| 9 |
LLM Pathway: Chapters 8–12 (Transformers, GPT, text generation).
Analytics: model refinement, error analysis.
RAG: multi-tool agents, evaluation frameworks, hallucination mitigation. |
|
| 10–11 |
Development Sprint 2 — evaluation, metrics, redesign if needed. |
|
| 12 |
Milestone 3 Workshop: evaluation, ablation studies, business value analysis. |
Milestone 3 Due |
| 13 |
Technical writing workshop, GitHub documentation structure, reproducibility rules. |
|
| 14 |
Presentation coaching; demo rehearsal; stakeholder framing. |
Final Report Due |
| 15 |
Final Presentations and Demonstrations |
Presentation |
Project Pathway Examples
1. Implementing a Large Language Model (PyTorch)
- Build a DCGAN to generate business-themed images (logos, store layouts, product mockups).
- Train a Transformer to summarize customer complaints or analyze call transcripts.
- Create a Hemingway-style or “brand-voice” text generator using GPT-style architectures.
- Fine-tune a diffusion model for product advertising imagery.
2. Applied AI Analytics Project
- Predict sales, churn, demand, or fraud using ML + LLM assisted features.
- Sentiment analysis from customer reviews using embeddings.
- Topic modeling with LLMs for unstructured text collections.
- Hybrid quantitative + generative insight report.
3. RAG / Business Process Automation
- Automate HR policy Q&A with a secure RAG chatbot.
- Automate customer service ticket routing using embeddings + classification.
- Build a procurement assistant that summarizes vendor contracts.
- Build a multi-step agent (approval workflows, research pipeline, report generation).
Final Deliverables Detail
- Working System / Model (GitHub repo with code + README)
- Dataset Documentation (sources, cleaning, schema)
- Evaluation (metrics, baselines, error analysis)
- Ethics & Bias Analysis (required section)
- Technical Report (15–25 pages)
- Live Demonstration (10–15 minutes)
Academic Integrity
All code must be original except for permitted libraries. LLMs may be used for assistance but must be cited. Using generative tools to produce full code or reports without attribution is considered academic misconduct.
|
