Pick a wonder, join its growth circle, build the project, and share your progress with the community.
“How do we find planets orbiting stars we can’t even see?”
Transit: a star’s brightness dips when a planet crosses in front of it.
Build it: Classify a star’s light curve over time to detect exoplanets (time-series ML).
“Why is one star blue and another red — and what does the color tell us?”
Wien’s law: hotter stars peak at bluer wavelengths (λ_max ∝ 1/T).
Build it: Train a classifier on stellar spectra to sort stars by type and temperature.
“How do you measure the distance to a star with a ruler you can never extend?”
Parallax: nearby stars shift against the background as Earth orbits (d = 1/p).
Build it: Build a CNN that classifies galaxy shapes — computer vision on the cosmos.
“Among a million points of light, how do you spot the one that just moved?”
Image differencing: subtract last frame from this one; what remains changed.
Build it: Detect asteroids & supernovae in image sequences by flagging what changed.
“Why does every flat map of the world lie about something?”
Gauss’s Theorema Egregium: a sphere can’t be flattened without distortion.
Build it: Classify satellite land-use (forest/city/water) with a CNN; track deforestation.
“Given only the shape of the land, can you predict where a flood will pool?”
Water follows steepest descent: flow accumulates along −∇(elevation).
Build it: Model flood risk from terrain + rainfall — where watersheds meet weather.
“Why do cities grow where they do — and who gets left off the map?”
Gravity model: interaction ∝ (Pᵢ·Pⱼ) / distance².
Build it: Optimize routes & accessibility across a region — the math behind a tourism guide.
“How do we “see” inside a planet we can never dig into?”
Locate a quake from the arrival-time gap between fast P-waves and slower S-waves.
Build it: Detect earthquakes in seismograms with anomaly detection; estimate the epicenter.
“Can you predict the next hour of rain from the last hour of sky?”
Rain forms when air cools to its dew point; storms advect with the wind.
Build it: A 6-hour precipitation nowcaster — the flagship “2002” weather project.
“Why can no one forecast the weather two weeks out — ever?”
Lorenz: tiny differences in starting conditions grow into wildly different outcomes.
Build it: Run ensemble forecasts (many slightly-different starts) and turn chaos into a probability.
“What makes the wind decide which way to blow?”
Wind flows down the pressure gradient (force ∝ −∇P).
Build it: Predict wind from pressure fields — a first taste of learning on vector fields.
“One is a mood, the other a personality — how do you tell them apart in the data?”
Weather is today’s value; climate is the trend across decades.
Build it: Detect a climate trend through noisy yearly data — climate data science that ships.
“Why does a shiny nail turn to orange dust in the rain?”
Redox: iron gives up electrons to oxygen (oxidation).
Build it: Predict material corrosion / battery life from conditions over time.
“Why is water bent and carbon dioxide straight — and why does it matter?”
Electron pairs repel, so geometry follows (VSEPR).
Build it: Predict molecular properties with a graph neural network on the molecule.
“How do chemists know a reaction will actually happen?”
A reaction proceeds when it lowers free energy (ΔG < 0).
Build it: Predict reaction feasibility and yield from the reactants.
“Out of millions of molecules, how do you find the one that heals?”
Similar structures tend to have similar effects (structure–activity).
Build it: Search molecular similarity with embeddings to shortlist drug candidates.
“How does a cell turn four letters into a living thing?”
The central dogma: DNA → RNA → protein, read in sequence.
Build it: Classify DNA sequences with a sequence model — biology as language.
“Why do you have your grandmother’s eyes?”
Mendelian probability: traits inherit with predictable odds.
Build it: Estimate genetic-risk probabilities with a Bayesian model.
“How does a chain of beads fold into a tiny machine — the same way every time?”
A protein settles into its lowest-energy 3-D shape.
Build it: Predict protein structure (a mini AlphaFold) by minimizing an energy.
“Can a computer spot a sick cell a doctor might miss?”
Cells differ in shape, texture and stain — features you can learn.
Build it: Classify microscope cell images with a CNN.
“How does a thin bridge hold up a thousand cars?”
Forces must balance — every push meets an equal pull (statics).
Build it: Predict and optimize where a structure will bear stress.
“With noisy sensors, how does a robot stay sure of its own position?”
Fuse motion and sensors by probability (Kalman / Bayes filter).
Build it: Build a localization filter that tracks position through noise.
“How does a machine hold an exact speed while the load keeps changing?”
Feedback control nudges toward the target and corrects the error (PID).
Build it: Tune an adaptive controller that learns to hold a setpoint.
“How do you hear one voice in a roaring crowd?”
Break a signal into frequencies (Fourier) and keep the ones that matter.
Build it: Denoise audio or vibration by filtering in the frequency domain.