2026

Michael Jiao

Statistics and Computer Science at Harvard. I spend my time figuring out where machine learning can be applied: physics, politics, and more.

Currently
CERN DeepLense · GSoC
And
Triangle Analytics · founding
Scroll
01now

CERN DeepLense

Apr 2026 → present · remote

Finding gravitational lenses in wide-field sky surveys, and getting the models to survive the jump from simulation to real telescope data. Google Summer of Code.

Triangle Analytics

Oct 2025 → present · Arlington, VA

Founding engineer. I build the forecasting models and the pipelines underneath them.

Harvard College

→ Dec 2028 · Cambridge, MA

Statistics, with a concurrent master's in computer science. Quantitative Traders, AI Safety Team, poker.

02writing

Cache You Lateraccepted

AdaptFM @ ICML 2026

Long-context models throw away tokens to save memory, then need them again. We put them back between turns — 91% retrieval where the baseline gets 24.5%, at the same cost.

When Do Looped Transformers Need More Loops?accepted

SciFM @ COLM 2026 · co-first author

Can you tell how long a model should think just by watching what it says? We built the benchmark to test it honestly. Mostly you can't — and the usual evidence that you can turns out to be noise.

Relevance Is Not Value

2026 · working

A forecasting agent has to choose what to read. Relevant isn't the same as useful, and it can often tell the difference before opening anything.

03built

PokerBot

2026

Reads a live table and solves for the right play in under a second.

Downstream

2026 · YHacks

Traces industrial PFAS discharge downstream to the fish you'd actually eat. 3,274 river segments, built in 16 hours.

GANaxy

2025

Sorted 300,000 galaxies and pulled out the 8,500 strange ones — mergers, starbursts, junk.

04record
Honorable Mention, Team USAIntl. Mathematical Modeling Challenge
Silver, FinalistBridgewater Forecasting
Bronze ×2USA Physics Olympiad
AIME Qualifier ×4Mathematical Association of America
IOAA TST, Team CanadaCanadian Astrophysics Olympiad