George Hu

About Me

Grade 8 · Shanghai

Engineer,
athlete, builder.

I'm George Hu (胡敬知) — a Grade 8 student at SMIC Private School, Shanghai. I like understanding how things work, then making them work better. That impulse has taken me from writing autonomous routines for a robot headed to the 2026 World Championship, to pulling an oar at dawn on the Huangpu, to shipping software projects nobody asked me to build.

The thread that connects all of it is the same: figure out the system, put in the reps, and build something real. Whether the system is a PID loop, a rowing stroke, or an argument in a debate round, the process never changes — and that's what I love about it.

FTC awards won

Years rowing competitively

World ranking — U19 relay

FTC World Championship

George Hu · 胡敬知

What I Do

Where I
invest my time.

Three commitments, one habit: show up, do the hard part first, and don't stop until the thing actually works.

Sep 2024 — Present · Team IK19859

FTC Robotics

Control systems lead, Team IK19859 — four FTC seasons. Every line of Java I write decides the robot's movement in a 30-second autonomous with no human on the controller. First Chinese team to reach Champion Alliance at California Invitational 2025. World Championship, Turkey, 2026.

Java PID Control Sensor Fusion Autonomous Systems Vice-Captain

Results & Awards

2026 FTC World Championship — Qualified, Turkey (Jul 2026)

Control Award — California Invitational 2025

Think Award — FTC Beijing Qualifier 2025–26

Innovate Award — Lobster Cup International 2025

Champion Alliance — Chongqing Qualifier & Lobster Cup

Finalist Alliance Runner-up — China Nationals

Team Impact

ScoutSelect, our alliance selection tool, has been adopted by teams beyond IK19859 — other FTC competitors have used it at tournaments in China. Writing code that teammates and other teams rely on during the pressure of competition changed how I think about software quality: it has to work when you're not there to fix it.

What failed

At China Nationals our autonomous routine failed in the first qualifying match — a sensor calibration drift we hadn't caught in practice. We had 8 minutes between matches to diagnose and patch it. We recovered, but the lesson held: test in conditions as close to competition as possible, not just in the workshop.

FTC robot trophy and IK19859 — competition floor

IK19859 — Competition Robot Overview

Click to explore Full Story

2023 — Present · Shanghai Youth Rowing Team

Competitive Rowing

5:30 AM, fog on the water. Three years, every season including winter. Sweep and sculling both. I film every practice, track every split, help set race strategy. World #6 at WRICH 2026. Nanjing 2026: Gold & Silver.

Nanjing 2026 WRICH 2026 SSSA 2026 HKRIC 2026 Sculls & Sweep Video Analysis

Race Results

🥇 Gold Univ. Men's Eights · 🥉 Bronze Men's Doubles — Shanghai Dream Regatta 2026

🥇 Gold Mixed Eight Sweep · 🥈 Silver Quad Sculls — Nanjing 2026

🥇 Champion U16 Male Relay — SSSA Indoor Rowing 2026

World #6 — WRICH 2026, U19 Mixed 5000m Relay

World #29 — WRICH 2026, U17 Men 1000m

#8 — HKRIC 2026, U14–16 Men 1000m (2000m: 7:28.9)

500m — 1:36 · 1000m — 3:33 · 2000m — 7:28 — HK Indoor PRs

Solo sculler at golden hour — training on open water

Golden hour — solo sculls

Mid-stroke intensity — crew at full power

Mid-stroke — full power

Click to explore Full Story

Ongoing · British Parliamentary Format

BP Debate

Fifteen minutes of prep. Then you stand, half a case ready, room waiting for you to fail. BP taught me to find the weak joint in an argument while building my own — under a timer, in front of judges who know the topic better. CISDAC Team Champion. Yale Global Rounds.

BP Format CISDAC SIDO 2024–26 GDC 2026 Yale Global Rounds

Competitions

Team Champion — CISDAC Open Division

Yale Global Rounds — World Scholar's Cup

SIDO — 3× Competitor (2024, 2025, 2026)

GDC 2026 — GuangDong Debate Championships

World Scholar's Cup — Gold · Debate, Writing & Scholar's Challenge

Shanghai IT 2026 — U14 Best Speaker · Schools Outrounds Finalist

George presenting at debate competition — centre stage with judges listening

CISDAC Open Division — Team Champion

Click to explore Full Story

Selected Work

What I've
built.

JouleRoute-LM Research Paper

Local LLMs always run full model depth — even for simple queries. I built adaptive layer routing by task complexity and energy budget, then tested it rigorously. Result: +14.43% task-success-per-joule over static baselines (Cohen's d = 7.49, p = 0.005). Independent research. Full paper on GitHub.

Python LLM Research Statistics Simulator

ScoutSelect

Three minutes to pick your FTC alliance. Raw scores lie — OPR via Gaussian elimination separates signal from noise. Bayesian shrinkage, 2,000 Monte Carlo sims per pairing, synergy scoring. Built by Team IK19859. Used at real competitions. Live at scoutselect.org.

Next.js 15 TypeScript React 19 Tailwind CSS v4 GraphQL

GHCountdown

I kept missing deadlines. What started as a weekend countdown grew through six months of daily use into a full local workspace: todos, timeline, time tracking, AI assistant, stats. No spec, no wireframes — every feature traces back to a wall I actually hit.

TypeScript React 19 Electron 41 Tailwind v4

7 More Builds ApexType · Nova SMP · Flight Network · Depowhiteboard · GHutils · DailyCLI · Velora Full write-ups →

Timeline

Competing,
building, &
leading.

A track record built across robotics, competitive sport, debate, and the classroom.

Sep 2024 — Present

Vice-Captain & Control Systems Lead

FTC Team IK19859

Built autonomous control systems in Java using PID tuning and sensor fusion. Won Control Award (California Invitational 2025), Think Award (Beijing Qualifier), Innovate Award (Lobster Cup International). China Nationals Finalist Alliance Runner-up. Qualified for 2026 World Championship in Turkey.

2023 — Present

Competitive Rower

Shanghai Youth Rowing Team

Training 6–8 hours per week. Nanjing 2026: Gold in Mixed Eight Sweep & Silver in Quad Sculls. SSSA 2026: U16 Male Relay Champion. WRICH 2026: World #6 in U19 Mixed 5000m Relay, World #29 in U17 Men 1000m. HKRIC 2026: #8 in U14–16 Men 1000m.

Ongoing · 5+ years

Debater & Team Champion

British Parliamentary & World Scholar's Cup

CISDAC Open Division Team Champion. World Scholar's Cup — Gold medals in Team Debate, Collaborative Writing & Scholar's Challenge. Qualified for Yale Global Rounds. SIDO competitor 2024, 2025 & 2026. GDC 2026. 10+ hrs/week of BP format training.

2026

AI Research Author

JouleRoute-LM · Independent Research

Authored original research paper on adaptive cross-layer inference control for local LLM workloads. Built Python simulator, 9-seed statistical pipeline, and paired hypothesis tests. Key result: +14.43% task-success-per-joule on chat workloads vs. static baseline (Cohen's d = 7.49, p = 0.005). Hosted at GitHub · N0v4ont0p.

2024–2025

Business Competitor

ASDAN Asia Business Challenge

Team Silver (2nd place) & Best Individual Presenter. Qualified for World Youth Economic Forum, Harvard Business Challenge, and ASDAN Financial Intelligence All-Star Forum.

2019–2020

Science Competitor

Early Science & Innovation

Started systematic science exploration at age 4. World Odyssey of the Mind — International Gold Medal (2019). ICW International Silver Medal (2020). Cross-disciplinary competition background that established a foundation for engineering and robotics.

2022 — Jun 2028

Student · Grade 8

SMIC Private School, Shanghai

School Captain (Primary School). Class Academic Monitor. Red Star Award recipient. English Week Chair & Emcee. Pudong District Children's Congress Delegate. MAP Growth: Reading 95th %ile, Language Usage 98–99th %ile.

April 2026 Debate

On being publicly wrong

BP debate is the only activity I've found where being wrong in front of an audience is the mechanism, not a side effect. The round doesn't end when you run out of arguments — it ends on a timer. You have to keep building even when you've spotted the hole in your own case. I've started applying this to debugging: don't stop the run just because something looks wrong. Finish the iteration.

March 2026 Rowing

What the 5:30 alarm actually teaches

Three years of rowing has produced one non-obvious lesson: the hardest part isn't the training session — it's the decision at 5:30 in the dark, before you're fully awake, to not negotiate with yourself. Once you're out the door the body figures it out. The skill being practiced is not endurance; it's pre-commitment. I think about that every time I'm tempted to skip a difficult debugging session.

February 2026 Research

Why I built a simulator before writing a paper

For JouleRoute I had a hypothesis but no dataset. The honest path was to build the environment that would generate the data — a Python simulator with controlled variables and real statistical tests. It took longer than just writing the idea down. But the result is that I actually know whether it works, instead of having an interesting argument about whether it should. Building the test before the claim changes how confident you need to be before you start.

Essays · Thought

How I
build.

Read all essays →

Engineer,
athlete, builder.

Numbers that
back it up.

Where I
invest my time.

What I've
built.

Competing,
building, &
leading.

How I
build.

Building people,
not just systems.

I'm always looking for the
next problem worth solving.

George Hu

Engineer,athlete, builder.

Numbers thatback it up.

Where Iinvest my time.

What I'vebuilt.

Competing,building, &leading.

How Ibuild.

Building people,not just systems.

I'm always looking for thenext problem worth solving.

Engineer,
athlete, builder.

Numbers that
back it up.

Where I
invest my time.

What I've
built.

Competing,
building, &
leading.

How I
build.

Building people,
not just systems.

I'm always looking for the
next problem worth solving.