INF 385T: Building Large Language Model Applications (UT Austin, Fall 2026)
by University of Texas at Austin (Jiaxin Pei)
A 16-week syllabus and reading list for building an agent-based LLM product and pitching it to investors.
Overview
INF 385T: Building Large Language Model Applications is a Fall 2026 course at the University of Texas at Austin taught by Jiaxin Pei. It runs for 16 weeks, meeting Mondays for three hours, and individual assignments in weeks 1 to 4 are followed by a team project in teams of three that makes up 60 percent of the grade. Early weeks survey frontier product areas, then cover how coding agents work, including agentic loops, tool use, function calling, structured outputs and MCP basics. Week 4 covers agents that act for users, with multi-step planning, error recovery, permissions, RAG, embeddings for memory and multi-agent orchestration. Later weeks cover generative engine optimization, demand and timing, cost optimization and deployment, including caching, model routing, latency, serving infrastructure and token budgeting, and then economics, moats and go-to-market. Teams present at a lightweight demo day, an internal demo day with beta testers and a final public demo day with investors. The public page links readings from Anthropic, OpenAI, Y Combinator and a16z, plus Lilian Weng's agent post. Exercises and code sit on Canvas, slides are not listed publicly, and the schedule is marked tentative, so self-study means following the readings and doing your own build. Enrollment needs graduate standing and instructor permission. Guest speakers include Xingyao Wang of OpenHands.
At a Glance
- Topic
- Agentic
- Level
- Intermediate
- Format
- Course
- Cost
- Free
- Duration
- 16 weeks, one 3-hour class per week (public syllabus and readings, self-paced)
- Provider
- University of Texas at Austin (Jiaxin Pei)
- Hands-on
- Yes — code/exercises
- Certificate
- None
What You’ll Learn
- ✓Explain how coding agents use agentic loops, tool use, function calling and structured outputs
- ✓Connect agents to external tools and data using MCP basics and multi-agent orchestration
- ✓Design personal agents that handle multi-step planning, error recovery, permissions and memory through embeddings
- ✓Reduce LLM application cost using caching, model routing, token budgeting and smaller models where suitable
- ✓Validate an AI product idea through competitor surveys, customer conversations and beta tests with real users
- ✓Apply generative engine optimization so that AI assistants are more likely to cite your product
- ✓Prepare a pitch deck, a demo and a Y Combinator application draft for your product
Highlights
- •The final grade weights a team-built product at 60 percent, with students shipping a publicly hosted prototype by week 3
- •Agent engineering sits next to distribution, pricing and go-to-market sessions in the same 16-week schedule
- •The confirmed speakers come from companies including OpenHands, Visa and Judgement Labs
- •Its reading list is public and links primary sources from Anthropic and OpenAI on tools, MCP and structured outputs
- •Self-learners get a partial course only: exercises and code stay on Canvas, and slides are not listed publicly
Who It’s For
Best For
- ✓Engineers planning to turn an agent or LLM prototype into a product
- ✓Founders with a technical background who want a week-by-week build and validation plan
- ✓Educators designing a project-based LLM applications course
Prerequisites
- •Comfort building software with coding agents, which the syllabus asks for instead of deep technical expertise
- •Graduate standing and instructor permission for UT Austin enrollment; neither is needed to read the public syllabus
FAQ
What is INF 385T: Building Large Language Model Applications (UT Austin, Fall 2026)?
INF 385T is a Fall 2026 UT Austin graduate course in which teams build and ship an LLM application, then pitch it at a public demo day. Its public syllabus suits engineers who want a reading-backed path through coding agents, MCP, RAG, memory, deployment cost and product validation for their own project.
Is INF 385T: Building Large Language Model Applications (UT Austin, Fall 2026) free?
INF 385T: Building Large Language Model Applications (UT Austin, Fall 2026) is free to access.
What level is INF 385T: Building Large Language Model Applications (UT Austin, Fall 2026) for?
INF 385T: Building Large Language Model Applications (UT Austin, Fall 2026) is aimed at a intermediate audience. Recommended background: Comfort building software with coding agents, which the syllabus asks for instead of deep technical expertise, Graduate standing and instructor permission for UT Austin enrollment; neither is needed to read the public syllabus.
How long does INF 385T: Building Large Language Model Applications (UT Austin, Fall 2026) take?
Expect roughly 16 weeks, one 3-hour class per week (public syllabus and readings, self-paced). Most learners work through it at their own pace.
What will I learn from INF 385T: Building Large Language Model Applications (UT Austin, Fall 2026)?
You'll learn: Explain how coding agents use agentic loops, tool use, function calling and structured outputs; Connect agents to external tools and data using MCP basics and multi-agent orchestration; Design personal agents that handle multi-step planning, error recovery, permissions and memory through embeddings; Reduce LLM application cost using caching, model routing, token budgeting and smaller models where suitable; Validate an AI product idea through competitor surveys, customer conversations and beta tests with real users; Apply generative engine optimization so that AI assistants are more likely to cite your product; Prepare a pitch deck, a demo and a Y Combinator application draft for your product.
Topics
Sources
This page was written from 2 sources, 1 on domains other than jiaxin-pei.github.io.