University of Southern California · Fall 2026
Capstone Course of the AI Minor @USC: Generative AI, NLP and GenAI Products
Welcome to TAC 459: Generative AI and Natural Language Processing, the capstone course of the USC Artificial Intelligence Minor at the University of Southern California. Co-instructed by Allen Bolourchi and Reza Jafarkhani, the course connects technical foundations, industry analysis, and hands-on AI product development through three complementary components.
Technical Foundations. Students explore how generative AI works and how frontier models are built, adapted, and evaluated. Topics include transformers and attention, prompting, RAG, efficient fine-tuning with LoRA and QLoRA, alignment, RLHF and PPO, model evaluation and benchmarks, and AI safety. Hands-on work connects these concepts to many use cases including conversational AI, text generation and summarization, machine translation, and multimodal applications.
AI Company Overview. Midway through the semester, teams investigate emerging AI companies backed by YC or elite VCs. Students analyze the products, underlying technology, customers, market opportunity, and business models, then present the company and a concrete GenAI use case to the class.
AI Capstone Product. Over nine weeks, teams apply class concepts across the AI product lifecycle: ideation, a shared minimum viable product, multiple prototypes, and a complete application built with APIs to frontier models. Students begin with course code and progressively use AI-assisted coding tools to build and refine their products. The course culminates in a working product demonstration and a startup-style launch pitch connecting the technical architecture with the user problem, product value, and business opportunity.
01
Technical core · Throughout the semester
Build the technical foundations of generative AI and natural language processing, from text processing and machine learning to transformers and frontier models. Connect these concepts with speech, alignment, and advanced generative AI applications.
02
Midsemester · Industry case study
Study an emerging AI company backed by YC or elite VCs through its product, underlying technology, customers, and market opportunity. Present the company and a concrete GenAI use case to the class.
03
Nine-week build · Ideas to a working product
Apply class concepts to a product built from the ground up. Progress from ideas and a shared MVP to multiple prototypes, then select, refine, and demonstrate a complete AI product with a frontier-model API feature, built with vibe coding using Codex, Claude, or Muse Code.
AI Research, Product and Program Management at Meta Superintelligence Labs, FAIR
Allen Bolourchi leads AI research and product at Meta Superintelligence Labs (MSL), Fundamental AI Research (FAIR), and is an Adjunct Professor of Generative AI and NLP at the USC Viterbi School of Engineering, where he directs the capstone of the USC AI Minor (TAC 459), which he designed from the ground up. At Meta, his work spans frontier models, agentic AI, world models, computer-use agents, and multimodal intelligence. He also co-leads the development of Meta's open-weight model Muse Glimmer and has co-authored research with Yann LeCun and other FAIR researchers. Previously, he led AI/ML personalization for the App Store and Apple Arcade at Apple.
Senior Lecturer of Technology and Applied Computing
Reza Jafarkhani is a Senior Lecturer of Technology and Applied Computing at the USC Viterbi School of Engineering, specializing in machine learning, artificial intelligence, and data analytics. He teaches all courses in the USC AI Minor, helping students build AI capabilities from foundational concepts to advanced applications. With extensive experience teaching students across the program, he connects technical theory with practical implementation and guides learners as they develop their AI skills. He co-instructs TAC 459 with Professor Allen Bolourchi, bringing this broad teaching experience to the course's technical foundations and hands-on AI product development.
Learning Assistant · Major: Business Administration, Applied Data Science
Dylan Chen is a USC student graduating in Spring 2027 with a B.S. in Business Administration, a Minor in AI Applications, and an M.S. in Applied Data Science. During his 2026 internship at Meta, he built Meta FinTech's offline feature platform behind personalized shopping and ad badges, serving all Meta users across a variety of domains, plus MLOps tooling and a semantic layer for the org.
Learning Assistant · Major: AI for Business
Arvin Duh is a USC student graduating in Spring 2027 with a B.S. in AI for Business. During his 2026 internship at Google, he built an agent-eval framework that cut evaluation time from 6 hours to 20 minutes, a 30-agent swarm mapping 3,000+ CFO Org profiles, a document parser, a rebuilt LoC tool, and a CL auto-review dashboard.
The AI Company Overview connects the technical material with the products and businesses being built around generative AI. Teams investigate an emerging company, explain why its technology matters, and use that analysis to inform their own product ideas.
Presentation: November 2, 2026 · 12:00 PM · In person
AI Company Overview presentation upload (presentation day only)
Form a team of six; four teams may have seven members. Select a company at the pre-seed, seed, Series A, or Series B stage with an emerging GenAI application. Y Combinator companies are preferred; other selections require professor approval. Discuss the choice with the professor before preparing the overview.
Cover the problem, product, generative AI technology, target customers, practical use case, differentiation, and market opportunity. Include one or two concrete GenAI examples so the class can understand how the product works and what makes it distinctive.
Prepare a 10-minute company overview and a two-minute Q&A, with every team member participating. Add one slide and a one-minute pitch for each proposed capstone product idea. Upload the presentation to Brightspace. Teams are also encouraged to contact a founder, executive, or employee; a 15–20 minute guest segment is optional if the company participates.
The AI Capstone Product turns course concepts into a working application over nine weeks. The same team moves from a problem and product idea to one shared minimum viable product (MVP), develops multiple prototypes, and selects a direction to refine into the complete product. Each product must use a frontier-model API for at least one feature.
Week 1 · , 2026
Use the same team as the AI Company Overview.
Add team members to the class sheet
No file upload required; update the class sheet and schedule your professor meeting.
Week 2 · , 2026
Develop initial product ideas grounded in class concepts and a clear user problem. Discuss possible applications and the value each idea could offer with teammates.
No upload required this week.
Week 3 · , 2026
Prepare one slide and a one-minute pitch for each proposed product idea, to present after the AI Company Overview on November 2.
No upload required this week.
Preparation milestone; no separate upload deadline this week.
Week 4 · , 2026
Present each product idea after the company overview, then define your product roadmap and the scope of one shared MVP. Meet with a learning assistant to enhance the roadmap and MVP.
AI Company Overview presentation upload (presentation day only)
Week 5 · , 2026
Use course code and class concepts to create one minimum viable product for the team. Establish the core user experience and technical foundation for the prototypes that follow.
Week 6 · , 2026
Improve the shared MVP and explore alternative product directions through multiple prototypes. Apply class concepts and incorporate a frontier-model API feature.
No upload required this week.
Week 7 · , 2026
Create prototypes in groups of two. For teams of seven, one group may have three members. Explore and compare different implementations before choosing a product direction.
No upload required this week.
Progress milestone; no separate submission deadline in the roadmap.
Week 8 · , 2026
Choose a prototype, enhance it, and bring the team’s work together into a complete application. Refine the chosen direction and prepare it for demonstration.
Week 9 · , 2026
Finalize your application and prepare a startup-style presentation connecting the user problem, product value, business opportunity, and technical architecture. Show the progression from the MVP through prototypes to the complete product. The roadmap schedules the in-person presentation for December 11 at 5:00 PM, subject to confirmation of exam conflicts. The presentation format will be shared.
AI Capstone Product presentation upload (presentation day only)
| Instructor / Learning Assistant | Day & Time | Location | Link |
|---|---|---|---|
| Reza Jafarkhani | Tuesday · 2:30 PM - 3:30 PM | RRB 211 / Zoom | Piazza |
| Allen Bolourchi | Announced on Piazza | Zoom | Piazza |
| Dylan Chen | Announced on Piazza | TBA | Piazza |
| Arvin Duh | Announced on Piazza | TBA | Piazza |
