Abhishek Bisht
Applied AI Research Engineer. Building and writing about the things that matter.
My notes on AI — rewritten from first principles, in code, with interactive demos.
One neuron. Then a network. Then GPT-2 — built by hand.
NLPHow language models work, misbehave, and get steered.
NLPThe internals most courses skip — efficiency, alignment, multimodal.
ArchitecturesWhat each landmark model actually invented — 2017 to now.
ArchitecturesFifty-one LLM diagrams. Zoom, search, compare side by side.
ArchitecturesOne codebase. Paper by paper. Watch the architecture grow.
Pre-TrainingInside the pretraining forward pass — every layer, every FLOP.
Deep LearningMLA, MoE, MTP, FP8 — rebuilt from first principles.
RL + ReasoningFrom chain-of-thought to GRPO. Teach a model to think.
InferenceServing at scale — paging, batching, speculation, VRAM.
SearchFrom BM25 to agents that search for themselves.
InterviewSeventy-five interview questions. Reveal, track, quiz yourself.
ResearchEvery paper cited here — searchable, tagged, one click away.
PythonPython that actually matters for ML — interactive, in the browser.
PythonA Python REPL in your tab. No install.
ArchivesEarly notes, 2018–2024. The path that led here.
CourseSix months. Ten phases. Applied AI research, end to end.