OIM3641 AI Driven APP Development
4 Advanced Elective Credits
Overview:
This course is designed for ambitious business students and aspiring entrepreneurs interested in moving beyond basic concepts in programming to explore launching AI-driven products. We treat Generative AI as a fundamental business tool, focusing on value creation and market delivery. Instead of deep theoretical computer science, we concentrate on the practical architecture, the rapid prototyping process, and the critical steps needed to transition a working script into a robust, deployable, and scalable application.
You will learn to develop AI systems that deliver tangible business value, grounded in your in team derived proprietary ideas. By the end of this course, you will have a production-ready application and the skills to iterate and adapt it to market demand.
Concepts Covered Include:
Product Ideation
Learn common frameworks for generating new product or feature ideas
Use storyboarding techniques to explain your idea to an audience
Learn how to use unified modeling language (UML) to clearly define, and keep your team talking about the same product
Vibe Coding for Rapid Prototyping
Every great AI product starts with a functional prototype. Vibe Coding enables entrepreneurs to start working with and iterating through versions of a product in hours or days instead of weeks or months. We focus on function-first development to quickly validate a product idea and gather early user feedback for refinement.
Speed-to-Market: Focus on core functionality using streamlined Python code to dramatically accelerate your development cycles and minimize costs before market validation.
Interactive Design: Use Streamlit—a powerful Python framework—to instantly turn data scripts and LLM outputs into professional grade, interactive web applications and dashboards, bypassing the need for complex front-end development.
Advanced Python Development
Building successful products requires more than just core LLM logic; it requires professional-grade development habits and tools.
Explore the basics and rationale of Object Oriented Programming (OOP)
Use Python with a virtual environment tool to isolate project dependencies, prevent dependency conflicts, and ensure your application runs identically across different developer machines and deployments
Learn version control with Git workflows for collaborative development, allowing you to track changes and roll back errors safely—a requirement for any professional software project.
Integrate cutting-edge, AI-powered IDE extensions (like code completion and generative code assistants) into your workflow to dramatically increase development speed and reduce the time spent debugging boilerplate code. This is the future of entrepreneurial coding.
Building Custom LLM Solutions (RAG)
General-purpose LLMs are powerful, but high-value, enterprise-grade solutions require specialized knowledge. You will master the Retrieval-Augmented Generation (RAG) architecture, the industry standard for creating reliable, grounded or context driven AI applications.
Proprietary Intelligence: Learn how to connect LLMs to internal documents, databases, or knowledge bases, ensuring your AI product provides answers that are relevant, accurate, and unique. Since your data is never shared with a public LLM, IP remains proprietary.
Data Pipeline: Understand the foundational data frameworks (like LlamaIndex) that manage complex data ingestion, indexing, and retrieval, turning unstructured text into a searchable asset for the LLM.
Advanced NLP for Business Insight
Beyond simple text generation, this module dives into advanced techniques for understanding and classifying user input and complex data—critical for automating business processes.
Hugging Face Ecosystem: Leverage the industry-leading Hugging Face Transformers library and its PyTorch backend. You'll use this toolkit to quickly access and deploy cutting-edge, pre-trained models.
Zero-Shot Classification (ZSC): Learn to classify customer feedback, support tickets, or market sentiment into categories without needing extensive, time-consuming labeling. ZSC is a massive accelerant for building immediate-value business systems, allowing you to instantly categorize and route new data streams.
Code Robustness and Refactoring to Production-Grade
The biggest differentiator between a hobby project and a funded startup is code resilience. This module ensures your application doesn't fail under pressure, protecting your user experience and reputation.
Elegant Error Handling: Implement robust Python error handling (try...except patterns) to gracefully manage common points of failure, such as missing API keys, failed data loads, or external API timeouts. This provides actionable feedback to the user, not just a crash.
Refactoring for Scale: Learn the essential process of Refactoring, moving your Vibe Code (temporary structures and hardcoding) into permanent, generalized functions and class abstractions. We will leverage Streamlit's built-in caching and error display tools to create a stable, responsive and reliable user experience.
Prerequisites: None