Lab · from scratch

Calling an API is easy. I wanted to know what happens underneath it.

So I sat down and implemented the modern NLP and LLM stack from first principles. Not to reinvent it. To understand it well enough that nothing about a model feels like magic.

Forty-five runnable notebooks, from a regex tokenizer to RLHF. I'm a PM now, but I don't want to manage what I can't reason about. Building these means I can argue about context windows, eval design, or why a model hallucinates from the implementation up, not from a blog post.

Code on GitHub ↗See the products →

45Notebooks, tokenizer to RLHF

232Cells

0Syntax errors

2Packs

01 Interpretability research · UNSW NLP

Can you tell what a model is thinking when its answer hides it?

A team research project for my NLP course at UNSW.

A language model can give the same answer to a clean prompt and a misspelled one. The output tells you nothing. So we asked a different question: does the model process them the same way inside? Think of it as a polygraph for a transformer. The words match; the internal signals might not.

We took pairs of prompts that produced the identical next token, then probed the residual stream to see if the clean and corrupted versions were still separable inside the network.

Step 1

How we tested it

Extracted activations with TransformerLens (hook_resid_post, last-token) across five models: GPT-2, GPT-2 Medium, Pythia-160M, Pythia-410M, GPT-Neo-125M.
Trained three probes at every layer (nearest-class-mean, logistic regression, MLP) with pair-aware 5-fold cross-validation to stop the probe memorising question topics.
Kept only output-invariant pairs: same top-1 next token, so any signal is purely internal.
Tested causality with activation patching: swapping clean activations in at the peak layer, and projecting out the mass-mean direction.

Step 2

What we found

All five models showed strong linear separability. Peak linear-probe accuracy 0.958 to 0.975; the parameter-free baseline stayed at 0.82 to 0.94, so it is a real geometric difference, not a probe artefact.
The signal peaked in the middle layers (8 to 12), where the model's representation of the input is most structured.
Causal patching moved it: swapping clean activations at layer 8 shifted output probability by +0.14, and erasing the mass-mean direction by -0.05. The direction is causal, not just correlated.
Ablation by corruption type: case changes were trivial to detect (1.0), spelling harder (0.88), punctuation near-invisible (0.5).

0.975Peak linear-probe accuracy

5Models probed

L8 to 12Where the signal peaks

+0.14Shift from causal patching

Mechanistic interpretabilityActivation / linear probingTransformerLensCausal interventions (activation patching)Representation geometry (mass-mean directions, PCA)Probing methodology (CV, leakage control)Multi-model evaluation (GPT-2 / Pythia / GPT-Neo)PyTorch · scikit-learn

The same idea is a practical model-monitoring tool: if internal activations track a shift the output hides, probing can flag it. That is the bridge from interpretability research to AI safety and product reliability.

Tokenization + regex + lexicon sentimentTF-IDF with Naive Bayes, logistic regression, SVMN-gram LMs, smoothing, perplexityWord2Vec / GloVe + semantic searchSequence tagging: HMM, Viterbi, CRF, BiLSTMNeural text classification in PyTorchAttention + a tiny Transformer, by handBERT / GPT / T5 with Hugging FacePretraining, fine-tuning, LoRA, SFT, DPOSummarization, QA, RAG, ROUGE + faithfulnessMachine translation + decoding + BLEUTokenization + regex + lexicon sentimentTF-IDF with Naive Bayes, logistic regression, SVMN-gram LMs, smoothing, perplexityWord2Vec / GloVe + semantic searchSequence tagging: HMM, Viterbi, CRF, BiLSTMNeural text classification in PyTorchAttention + a tiny Transformer, by handBERT / GPT / T5 with Hugging FacePretraining, fine-tuning, LoRA, SFT, DPOSummarization, QA, RAG, ROUGE + faithfulnessMachine translation + decoding + BLEU

Product evals: kappa, PII redaction, monitoringTokenizers, embeddings, RoPE / ALiBi, attention, MHATransformer blocks, mini-former, sampling, speculative decodingKV cache, MQA / GQA / MLA, long context, FlashAttentionMixture of experts, sparse + state-space models, diffusion LMsSynthetic data, scaling laws, SFT / DPO / RLHF / GRPOQuantization + serving stacksEval harnesses, RAG, tool use + agentsVision-language adapters, interpretabilityRed-teaming + a capstone that wires it togetherProduct evals: kappa, PII redaction, monitoringTokenizers, embeddings, RoPE / ALiBi, attention, MHATransformer blocks, mini-former, sampling, speculative decodingKV cache, MQA / GQA / MLA, long context, FlashAttentionMixture of experts, sparse + state-space models, diffusion LMsSynthetic data, scaling laws, SFT / DPO / RLHF / GRPOQuantization + serving stacksEval harnesses, RAG, tool use + agentsVision-language adapters, interpretabilityRed-teaming + a capstone that wires it together

02 The notebooks

Two packs, 45 notebooks.

1313 notebooks

NLP implementation

The arc from classic NLP to transformers, one technique at a time. Build each thing, run it on small data, see why the next thing exists.

Open on GitHub ↗

3232 notebooks

Advanced LLM systems

The systems layer. How a modern LLM is actually built, served, and kept honest.

Tokenizers, embeddings, RoPE / ALiBi, attention, MHATransformer blocks, mini-former, sampling, speculative decodingKV cache, MQA / GQA / MLA, long context, FlashAttentionMixture of experts, sparse + state-space models, diffusion LMsSynthetic data, scaling laws, SFT / DPO / RLHF / GRPOQuantization + serving stacksEval harnesses, RAG, tool use + agentsVision-language adapters, interpretabilityRed-teaming + a capstone that wires it together

Open on GitHub ↗

Verification

Does it actually run

Every code cell parses clean (45 notebooks, 232 cells, zero syntax errors). The classic notebooks run end to end on CPU with toy data. The heavy pieces (FlashAttention, vLLM serving, full RLHF) ship as teaching versions plus production templates.