Skip to content

Latest commit

 

History

History
291 lines (232 loc) · 101 KB

File metadata and controls

291 lines (232 loc) · 101 KB

Synapsa Commons › Track lessons

Track lessons

Lessons that belong to a subject track rather than to one industry programme. Each is a notebook you do, with the same exercises, checks and hints as every other lesson here, and each opens in Colab, Kaggle or Binder from the links beside it, or in Codespaces or your own Jupyter as the README describes; the worked solution and the full rubric come with enrolment on Synapsa.

New here? Take the beginner path, or pick a track.

Orientation track lessons Data and statistics track lessons Deep learning foundations track lessons Language models track lessons
Post-training track lessons Inference systems track lessons Retrieval and RAG track lessons Agents and protocols track lessons
AI engineering track lessons Evaluation and governance track lessons AI for science track lessons

Start from zero

Begin with some Python and take these in order; each needs only the lessons before it.

  1. The 8 GB track · ~40 minutes · Python · no lesson to do first
    Build a profiler that measures what any piece of Python really costs in time and memory, rank four implementations of one task on the numbers, and reproduce the compute-tier check every lesson here has to pass.
    Open The 8 GB track in Colab Open The 8 GB track in Kaggle Open The 8 GB track in Binder
  2. Byte-level BPE from scratch · ~60 minutes · Python · builds on The 8 GB track
    Build byte-level byte pair encoding from scratch, and measure the token tax it imposes on five writing systems.
    Open Byte-level BPE from scratch in Colab Open Byte-level BPE from scratch in Kaggle Open Byte-level BPE from scratch in Binder
  3. The CPU slice of nanolm · ~75 minutes · Python · builds on The 8 GB track
    Train a character-level language model on a CPU with numpy alone, its backward pass derived by hand and checked against finite differences.
    Open The CPU slice of nanolm in Colab Open The CPU slice of nanolm in Kaggle Open The CPU slice of nanolm in Binder
  4. Retrieval from scratch · ~70 minutes · Python · builds on The 8 GB track
    Build BM25, a dense retriever with no external model and hybrid fusion, scored by an evaluation harness whose intervals say when one retriever is really better.
    Open Retrieval from scratch in Colab Open Retrieval from scratch in Kaggle Open Retrieval from scratch in Binder
  5. The agent loop from scratch · ~55 minutes · Python · builds on The 8 GB track
    Build an agent harness with validated tools, budgets, retries and traces, and measure how per-step reliability compounds over long tasks.
    Open The agent loop from scratch in Colab Open The agent loop from scratch in Kaggle Open The agent loop from scratch in Binder

Orientation

4 lessons

Start from nothing, on a laptop. Measure what code costs on an 8 GB machine, keep a model registry honest through a model swap, and write one client for three backends.

  • The 8 GB track · ~40 minutes · Python · no lesson to do first
    Build a profiler that measures what any piece of Python really costs in time and memory, rank four implementations of one task on the numbers, and reproduce the compute-tier check every lesson here has to pass.
    Open The 8 GB track in Colab Open The 8 GB track in Kaggle Open The 8 GB track in Binder
  • A model registry that survives a swap · ~40 minutes · Python · builds on The 8 GB track
    Build a role-based model registry, a retirement detector run against a dated snapshot of a provider's model pages, and a golden test that survives a model swap.
    Open A model registry that survives a swap in Colab Open A model registry that survives a swap in Kaggle Open A model registry that survives a swap in Binder
  • One client, three backends · ~50 minutes · Python · builds on The 8 GB track
    Build the normalising half of a provider-agnostic LLM client, which reads tool calls, reasoning and JSON out of whatever shape a backend returns, and a compatibility suite that shows what three stand-in backends keep, rename, drop or reject.
    Open One client, three backends in Colab Open One client, three backends in Kaggle Open One client, three backends in Binder
  • Living on free tiers · ~90 minutes · Python · builds on The 8 GB track
    Build a token-bucket limiter, jittered retries, a pacer and a capacity planner for a class on free API tiers, and measure how many learners finish, how many requests are wasted and what a quota cut does
    Open Living on free tiers in Colab Open Living on free tiers in Kaggle Open Living on free tiers in Binder

Data and statistics

2 lessons

Statistics by simulation: sampling distributions, bootstrap intervals, permutation tests and power, the habits every later lesson leans on.

  • Statistics by simulation · ~90 minutes · Python · builds on The 8 GB track
    Build a simulation toolkit for experiment analysis -- sampling distributions, bootstrap intervals, a permutation test and power calculator -- and audit how peeking and testing twenty metrics inflate false positives
    Open Statistics by simulation in Colab Open Statistics by simulation in Kaggle Open Statistics by simulation in Binder
  • How models lie · ~50 minutes · Python · builds on The 8 GB track and Statistics by simulation
    Build a leakage clinic — plant five leaks (target, preprocessing, duplicate, group and temporal) in a seeded customer book, measure how much each inflates cross-validated AUC against an oracle ceiling, and write the split, feature and pipeline checks that catch all five
    Open How models lie in Colab Open How models lie in Kaggle Open How models lie in Binder

Deep learning foundations

2 lessons

Build automatic differentiation from scratch, a scalar engine and a tensor engine, and the gradient check that keeps both honest.

  • Autodiff from scratch · ~75 minutes · Python · builds on The 8 GB track
    Build a scalar and a tensor reverse-mode autodiff engine, plus a forward-mode one, and hold every gradient to a finite-difference check
    Open Autodiff from scratch in Colab Open Autodiff from scratch in Kaggle Open Autodiff from scratch in Binder
  • PyTorch core idioms · ~90 minutes · Python · builds on Autodiff from scratch
    Write a PyTorch training loop with a checkpoint that survives a restart, and measure what each missing piece of state costs
    Open PyTorch core idioms in Colab Open PyTorch core idioms in Kaggle Open PyTorch core idioms in Binder

Language models

3 lessons

Tokenisers from scratch, the BPE merge loop in C++, and a language model trained on a CPU with numpy alone.

Post-training

8 lessons

Post-training from scratch: GRPO in numpy on a tiny model, and the traps that make a reward mislead you, measured.

Inference systems

4 lessons

Serve a language model on modest hardware: a KV cache, paging and continuous batching on a CPU, and what each one buys.

Retrieval and RAG

4 lessons

Build search and RAG you can measure: retrieval, chunking, the context budget, reranking and agent memory.

Agents and protocols

10 lessons

Build AI agents that hold up: the agent loop from scratch, MCP and A2A from the wire up, evaluation with intervals, and constrained decoding.

AI engineering

4 lessons

Run agents like production software, with OpenTelemetry traces and a cost-per-task report rebuilt from the traces alone.

Evaluation and governance

8 lessons

Measure models honestly: an EU AI Act evidence pack, evaluation across languages, and a mini arena whose ratings carry intervals.

AI for science

2 lessons

Do AI for science on real data: find the first gravitational wave in the raw LIGO data, and model rainfall to runoff honestly.

  • Find GW150914 in the raw LIGO data · ~50 minutes · Python · builds on The 8 GB track
    Find the first gravitational wave ever detected in the raw LIGO data, on a laptop: noise spectra, whitening, a chirp template and matched filtering, from scratch.
    Open Find GW150914 in the raw LIGO data in Colab Open Find GW150914 in the raw LIGO data in Kaggle Open Find GW150914 in the raw LIGO data in Binder
  • Rainfall to runoff, honestly · ~60 minutes · Python · builds on The 8 GB track
    Build a four-parameter rainfall-runoff model, a calibrator and a lagged-rain baseline, and measure the hold-out, split and flow-band traps on two real Australian catchments.
    Open Rainfall to runoff, honestly in Colab Open Rainfall to runoff, honestly in Kaggle Open Rainfall to runoff, honestly in Binder

Synapsa Commons · © 2026 RealAI · licensed under CC BY-NC-SA 4.0. You may use this material to learn and to teach, and copy, fork, share and adapt it. You must credit "Synapsa Commons by RealAI" with a link to https://github.com/TarrySingh/Artificial-Intelligence-Deep-Learning-Machine-Learning-Tutorials, say what you changed, and share adaptations under the same licence. You may not use it, or anything adapted from it, for commercial advantage or payment; for a commercial licence, contact RealAI. The full summary is in NOTICE; the licence's legal code governs.