Learn AI Create the Future — with Pranay
Main site

All sections

The whole curriculum, grouped into 14 areas. Start at the top if you are new. If you came here for one thing, filter to its area or search — everything is on this page or one click from it.

Every lesson here is written by .

14 areas 117 sections 1,162 of 1,167 lessons published 10 new lessons every day

Showing AI in the Real World — 12 sections, 111 lessons. Show all 14 areas

AI in the Real World

What AI actually looks like in hospitals, banks, shops, factories, farms, schools and laboratories.

12 sections 111 of 111 lessons published

AI in Healthcare

What changes when the model's output touches a patient — messy hospital data, rare events, and mistakes that hurt real people.

8 published

  1. What hospital data actually looks like
  2. Irregular clinical time series
  3. Predicting patient deterioration
  4. Label leakage in clinical models
  5. Survival analysis and censored outcomes
  6. False alarms and alert fatigue
  7. Validating a model before it touches patients
  8. When medical AI has caused harm

Medical Imaging AI

Scans are not photographs. Different file formats, different labels, different ways to be quietly wrong.

9 published

  1. DICOM, windowing and image intensity
  2. Segmenting 3D scans
  3. When the radiologists disagree
  4. Screening: one positive in a thousand scans
  5. Gigapixel pathology slides
  6. Multiple instance learning
  7. Why your model fails at the next hospital
  8. Generating radiology reports
  9. Reconstructing scans with learned priors

AI in Finance and Fraud

Money data is adversarial, regulated, and full of traps that make broken models look brilliant.

9 published

  1. Point-in-time financial data
  2. Backtesting without fooling yourself
  3. Credit scoring
  4. Reject inference
  5. Explaining a credit decision to a regulator
  6. How a real fraud system is built
  7. Finding fraud rings with graphs
  8. Anti-money-laundering monitoring
  9. Model risk management

Retail, Demand and Supply Chain

Forecasting things you sell and moving them to people — where the forecast is only useful if it turns into a decision.

12 published

  1. Retail demand data
  2. You measure sales, not demand
  3. Intermittent demand and mostly-zero series
  4. Making store, region and national forecasts add up
  5. Turning a forecast into an order quantity
  6. Promotions, cannibalisation and halo
  7. Price elasticity and dynamic pricing
  8. Forecasting a product with no history
  9. Predict then optimise
  10. Predicting arrival times
  11. Snapping noisy GPS to roads
  12. Messy addresses and geocoding

Manufacturing and Predictive Maintenance

Factories generate enormous sensor data and almost no failures — which flips how you build and prove a model.

10 published

  1. Industrial sensor and historian data
  2. Predictive maintenance
  3. Remaining useful life
  4. Building a model from four failures
  5. Vibration analysis for rotating machines
  6. Visual defect inspection
  7. Soft sensors
  8. Control charts versus machine learning
  9. Yield analysis and wafer maps
  10. Deploying a model inside a factory

Robotics, Control and Autonomy

AI that moves physical objects, where a mistake is not a wrong prediction but a collision.

10 published

  1. Robot sensors and time synchronisation
  2. Sensor fusion and the Kalman filter
  3. PID control, and when to use it instead
  4. Sim-to-real transfer
  5. Imitation learning and behaviour cloning
  6. Grasping and manipulation
  7. Predicting what people and cars will do next
  8. Inside an autonomous driving stack
  9. Vision-language-action models
  10. Safety in physical AI systems

Climate and Energy AI

Forecasting weather, sunshine and electricity, where being 2% wrong has a published cost.

7 published

  1. Weather and climate data formats
  2. Machine-learned weather forecasting
  3. Downscaling climate projections
  4. Forecasting solar and wind output
  5. Electricity load forecasting
  6. Battery dispatch and grid decisions
  7. Modelling wildfire and flood risk

Satellite and Agricultural AI

Models that look down from space at fields, forests and cities, and the pixels they have to work with.

8 published

  1. Working with satellite imagery
  2. Detecting change from space
  3. NDVI and vegetation indices
  4. Predicting crop yield
  5. Why crop disease models fail in the field
  6. Weed detection and spot spraying
  7. AI for smallholder farms
  8. GeoTIFF tiling and georeferencing

AI in Cybersecurity

The only domain where a human opponent studies your model and changes their behaviour to beat it.

9 published

  1. Security telemetry
  2. Triaging alerts at scale
  3. Malware classification
  4. Your opponent updates too
  5. Detecting phishing from URLs and headers
  6. Parsing logs before you can model them
  7. Data poisoning and backdoors
  8. Stealing a model through its API
  9. Securing the ML supply chain

AI in Education

Modelling how people learn, where the metric that matters is learning, not clicks.

10 published

  1. What learning data looks like
  2. Knowledge tracing
  3. Item response theory
  4. Adaptive testing
  5. Spaced repetition scheduling
  6. Automated essay scoring
  7. Building an LLM tutor
  8. Predicting which students will struggle
  9. Why AI-text detectors do not work
  10. Measuring learning, not engagement

AI for Accessibility

AI that removes barriers for disabled people — the highest-value applications, built on the thinnest datasets.

10 published

  1. Generating alt text that is actually useful
  2. Live captioning
  3. Speech recognition for atypical speech
  4. Sign language recognition
  5. Reading the world aloud
  6. Navigation aids for blind users
  7. Predictive text for AAC devices
  8. Plain-language rewriting
  9. Disability and dataset bias
  10. Evaluating assistive AI with real users

AI for Science and Engineering

Using models to replace expensive simulations and experiments, where a number without an error bar is useless.

9 published

  1. Surrogate models for simulation
  2. Physics-informed neural networks
  3. Neural operators
  4. Machine-learned interatomic potentials
  5. Predicting molecular properties
  6. Protein structure prediction
  7. Generative design of molecules
  8. Machine learning on genomic data
  9. Choosing the next experiment to run

Other ways to use this site

Learning paths

An ordered route through the lessons for one job: ML developer, AI engineer, computer vision engineer.

Projects

Build real things with full code, a dataset and a deployment guide.

AI glossary

Every term you keep seeing, defined in one plain sentence first.

Error database

Paste the error you got. Find out what it means and how to fix it.