Pre-Med
Build clinical AI.
I built an AI agent that takes patient intake, summarizes for the attending, and flags red flags.
Open now: College Readiness Coaching · Parent AI Workshop · Summer Institute planned for 2028, not yet enrolling
Senior engineer-led · Co-taught by working professionals · Selective admissions
Founded by Jinwoong Lee — UC Berkeley ’18 · 8 years of software at Amazon & Ring · 80+ students placed at top universities
Where the future doctor, lawyer, founder, engineer, and analyst learn to build the AI their profession will run on.
Students build a small AI system, without writing code, that organizes their studying. They still do the studying. The founder coaches each family directly, anchored to a real application date.
A study system the student runs1:1 College readiness coaching
Prefer to read first? Read the parent workshop →
AI College Prep · By the Numbers
A program built — every detail — around outcomes.
5
Pre-Professional Tracks
8:1
Student-to-Mentor Ratio
3 wks
Summer Intensive
90 hrs
Contact Instruction
100%
Practitioner-Taught, By Design
40-60%
Target Admit Rate
1:1
Admissions Coaching
8 yrs
Production Engineering Lead
2028
Inaugural Cohort
The Problem
Other AI summer camps teach 2018-era machine learning. We teach the agentic AI your child’s profession will run on.
Look closely at the syllabus of any other AI summer program. Pneumonia classifiers. Iris-flower regressions. Twenty-five-hour group classes taught by graduate students. Useful, maybe — but not what your child will actually be using day-to-day in medicine, law, finance, or engineering.
Other AI summer camps
A capstone you have seen many times before
- Teach 2018-era ML pneumonia classifiers
- 25-hour group classes
- Grad student instructors
- Generic projects, no domain depth
AI College Prep Academy
A capstone built around your child’s actual future profession
- Build agentic AI for your specific profession
- 8:1 cohort with working practitioners
- Senior engineer + working professional co-teach
- Real domain projects (clinical intake, contract review, financial analysis)
What we actually teach — in 30 seconds
Nothing vanishes. Every miss becomes the next session.
Practice_Section_RW_Module2.pdf
27 questions · graded against the student’s own error log
Q1 – Q27 · every question, graded
missed — into the log shaky — timed re-run solid — move on
error-log.md — the student’s own file
- Q7Comma splice vs. period — picked B, answer D. I treat "however" like "and."
- Q19Transition question — chose the contrast word out of habit, not from the sentence.
- Q3 · Q12 · Q22 · Q26Right, but slow or second-guessed — flagged for a timed re-run, not for drilling.
The instruction — written by the student, readable by you
Build tonight’s 20-minute drill from these six. One at a time, hardest first — and don’t show me the answer until I commit to one.
Scripted demonstration on a synthetic section — no student data, and no model that knows or remembers anyone. The error log is the student’s own file; it is the continuity.
AI helpers, no code
Your student builds a study coach. You build the family’s college concierge.
What the industry calls an AI agent is a job plus a written rule. The student’s version coaches instead of answering. The parent’s version keeps the dates and drafts the emails you send. Both are built in plain English and both are yours to keep. Each run starts from the page and material the family supplies; a person checks every output.
Your student, with you watching
The study coach
Asks the tool to file each practice-test miss under the concept it tested and ask what the student was thinking. The instruction tells it not to write a sentence the student hands in, and the student checks that it followed the rule.
You, the parent
The family college concierge
Starts from the dated list you supply: test dates, application deadlines, scholarship cutoffs, recommendation asks. It drafts; you check and send.
A school’s office and teaching staff
Office and classroom helpers
Sorts a week of parent questions into the ones that need a person and drafts replies to the rest for a person to review. No student records go in, and we do not write a school’s AI policy. A staff session is not a product yet. If it would be useful, tell us what it would need to cover.
What your child will build
For the future doctor, lawyer, founder, engineer, and analyst.
Five tracks. Five real agents. Each one ships to a personal URL by demo day, ready for your child to demo in their next college interview.
Watch one work
This is what “build a real agent” means.
Not a chatbot transcript — a working pipeline. The pre-med capstone takes a chief complaint, runs a structured follow-up, retrieves clinical guidelines, drafts the note, and flags what a physician must review. Your child builds this, step by step, and ships it to a personal URL.
Clinical Intake Agent
Pre-Med capstone · illustrative run
Receives the patient’s concern in plain language
Synthetic example: persistent cough, nine days
Asks five structured follow-up questions
Ordered by urgency
Retrieves the relevant clinical guidelines
Three sources matched
Drafts the History & Physical note
Citations attached
Flags one finding for the physician to review first
Draft ready for physician review
Illustrative run with a synthetic patient — this is the capstone a pre-med student builds and ships.
Pre-Law
Build legal AI.
I built an AI agent that reviews contracts and flags non-standard clauses.
Pre-Finance / Business
Build analyst AI.
I built an AI agent that pulls 10-Ks, computes financial ratios, and surfaces management's framing changes.
Pre-Engineering
Build hardware AI.
I built an AI agent that suggests CAD revisions based on tolerance constraints.
Pre-Entrepreneurship
Build founder AI.
I built an AI agent that runs my customer-discovery pipeline end-to-end.
The Curriculum
Three weeks. Five tracks. One agentic AI portfolio.
01
Foundations (shared)
All 5 tracks together. Build literacy with prompt engineering, RAG, and agent orchestration before splitting into domain work.
- Prompt engineering & evaluation harnesses
- Retrieval-augmented generation (vector DBs, chunking, hybrid search)
- Agent loops, tool use, and the Model Context Protocol
- Reading and debugging code with AI as a co-pilot
02
Domain Deep Dive (split by track)
Tracks split. Each cohort works with their track lead on domain-specific data, workflows, and the hard problems unique to that profession.
- Track-specific datasets and evaluation rubrics
- Domain-aware prompts, system instructions, and guardrails
- Compliance and ethics — medical, legal, financial framing
- Capstone scoping and architecture review with the track lead
03
Build, Ship & College Admissions Coaching
Build the capstone, ship it to a public URL, and layer 1:1 college admissions coaching on top throughout the week.
- Capstone build sprints with daily code review
- Demo day rehearsal — slides, talk track, live demo polish
- 1:1 college essay coaching (3 sessions)
- Mock interview, narrative architecture, and Common App essay draft
The Admissions Wedge
Three coaching sessions. One Common App essay draft. Built around your child’s project.
Week 3 includes 3 one-on-one essay coaching sessions, a mock interview, a narrative architecture session, and a finished Common App essay draft built around your child’s capstone project. We have not found another AI summer program that includes this.
- 3 one-on-one essay coaching sessions
- Mock interview with feedback
- Narrative architecture session
- Finished Common App essay draft
Founder · Senior Engineer
Jinwoong Lee
Founder, AI College Prep Academy · Senior software engineer · 8 years building software
Jinwoong is a 2018 UC Berkeley graduate who has worked the rare double track that makes AI College Prep Academy possible: eight years building software at companies that include Amazon and Ring (he is currently a senior software engineer), alongside eleven years running his own private education practice that began while he was still an undergraduate. He has also founded and shipped AI software for non-technical owners in a privacy-regulated healthcare industry — work that sharpened the exact skill the AI College Prep Academy curriculum is built around: teaching people who aren't engineers to use, evaluate, and ship AI in real-world work. Across more than eighty mentees he has placed students at Harvard, Stanford, Princeton, Columbia, Brown, Dartmouth, Cornell, UC Berkeley, UCLA, Northwestern, Georgia Tech, and other top-twenty universities, and his students have been admitted to selective summer programs including the Wharton M&TSI and have been district winners of the Congressional App Challenge.
Each of our 5 tracks will be co-taught by a working professional in that field — a medical resident, a law fellow, a finance analyst, a practicing engineer, or a startup founder — faculty we are recruiting from Northwestern Feinberg, UChicago Pritzker Law, Booth / Kellogg, McCormick School of Engineering, and the Chicago tech founder ecosystem.
8 yrs
Production software engineering
80+
Students placed at top universities
11 yrs
Private education practice
The 8:1 Cohort Promise
Eight students per track. Daily 1:1 attention.
8 students per track
Each track is intentionally small. With eight students per track, your child gets daily 1:1 attention from the mentor team — a staffing ratio built into the program’s design, not an add-on.
Our position on AI
The student does the thinking.
Every family asks a version of the same question before they ask about price: is this cheating? We answer it in public, name the things we refuse to do, and say so when the honest answer is that nobody knows yet.
- The student does the thinkingAI can explain a missed problem, quiz a student on a weak skill, or reorganize messy notes. It does not produce work the student turns in as their own. If a workflow only works when the student stops thinking, we do not teach it.
- Every answer gets verifiedWe teach the check, not just the prompt. An AI explanation is confirmed against the textbook, the scoring guide, or the original source before it becomes something a student studies from. Students should leave more skeptical of a confident wrong answer than when they started.
- No black boxesWe show the instruction that produced each output. A student who can read and edit their own prompt owns the system and can repair it. A student handed a finished bot owns nothing and learns nothing when it breaks.
What your child walks away with
Eight outcomes. Every student. Every track.
Working AI agent
Live at a personal URL — your child can demo it in college interviews, internships, and parent-teacher nights.
Common App essay draft
Finished draft built around the capstone, ready for fall coaching and final revisions with your senior-year counselor.
Project portfolio
Documented codebase, demo video, and architecture writeup — every artifact a college admissions reader could verify.
Reusable AI workflows
Productivity workflows for SAT/ACT prep, research papers, and college essay iteration — usable for years after the program.
Peer cohort
8 students per track. A lifelong professional network of similarly motivated, similarly serious peers.
Practitioner mentor relationship
Direct mentor in your child's target field — for advice, references, and introductions long after the program ends.
Letter of recommendation
When performance warrants. Personal, specific, and written by a working professional — not a templated form letter.
Mentor introductions
Direct intros to working professionals in your child's target field — beyond the assigned mentor, into the broader practitioner network we work with.
How families work with us
The path most families take.
You don’t have to go in order. Most parents start at the seminar; many start at 1:1. The 4 stages are designed so each one ladders naturally into the next — same founder, same engineer, no relationship reset between programs.
Format & Pricing
Two ways to work with us. Both explained here.
An ongoing monthly retainer for families who want the work carried month to month, or a one-time sprint for families who want a written plan and a decision. What each includes is laid out below; pricing is confirmed on your consultation.
Recurring · Live
College Readiness Coaching
By consultation
billed monthly · month to month · six seats
Founder-led · anchored to the next milestone
- Every session with the founder — not a junior tutor
- Each month opens on a real date: a test, an AP week, a deadline
- A study system your student builds and keeps running without us
- A monthly parent report you can read in five minutes
- Honest reporting, including the parts that are not working
- No build fee, no onboarding fee, no add-ons
Six seats, because that is how many one person can do this way.
One-time
Diagnostic Sprint
By consultation
one-time engagement · paid discovery
For a decision, not a retainer
- A written assessment of the current workflow and where it breaks
- A prioritized plan with owners and dates that you keep
- A clear recommendation — including "you do not need us"
- No obligation to continue into the retainer afterward
One-time work. We never report it as recurring revenue.
Questions about cost or fit, or not sure which one you need?
Talk to admissions
Demo Day · How the three weeks end
Eight students per track. Eight working AI agents, presented to the room.
Illustration — the inaugural in-person cohort runs Summer 2028.
Selective by design
Ready to give your child the agentic-AI advantage?
The Summer Institute is planned for 2028 and is not enrolling; no application is being accepted yet. What is running now is founder-led college readiness coaching, and a consult is where every family starts.
“We do not yet have AI College Prep alumni data — by intent, not omission.”
— Jinwoong “Peter” Lee, founder. No guarantees, no implied admit promises — outcomes get published here when they are real.