akillness/ai-research-skills

1

nanogpt Educational GPT implementation in ~300 lines. Reproduces GPT-2 (124M) on OpenWebText. Clean, hackable code for learni…

akillness/ai-research-skills

1

2

nemo-curator GPU-accelerated data curation for LLM training. Supports text/image/video/audio. Features fuzzy deduplication (16× fa…

akillness/ai-research-skills

1

3

nemo-evaluator-sdk Evaluates LLMs across 100+ benchmarks from 18+ harnesses (MMLU, HumanEval, GSM8K, safety, VLM) with multi-backend exe…

akillness/ai-research-skills

1

4

nemo-guardrails NVIDIA's runtime safety framework for LLM applications. Features jailbreak detection, input/output validation, fact-c…

akillness/ai-research-skills

1

5

nnsight-remote-interpretability Provides guidance for interpreting and manipulating neural network internals using nnsight with optional NDIF remote …

akillness/ai-research-skills

1

6

openrlhf-training High-performance RLHF framework with Ray+vLLM acceleration. Use for PPO, GRPO, RLOO, DPO training of large models (7B…

akillness/ai-research-skills

1

7

optimizing-attention-flash Optimizes transformer attention with Flash Attention for 2-4x speedup and 10-20x memory reduction. Use when training/…

akillness/ai-research-skills

1

8

outlines Guarantee valid JSON/XML/code structure during generation, use Pydantic models for type-safe outputs, support local m…

akillness/ai-research-skills

1

9

peft-fine-tuning Parameter-efficient fine-tuning for LLMs using LoRA, QLoRA, and 25+ methods. Use when fine-tuning large models (7B-70…

akillness/ai-research-skills

1

10

phoenix-observability Open-source AI observability platform for LLM tracing, evaluation, and monitoring. Use when debugging LLM application…

akillness/ai-research-skills

1

11

pinecone Managed vector database for production AI applications. Fully managed, auto-scaling, with hybrid search (dense + spar…

akillness/ai-research-skills

1

12

prompt-guard Meta's 86M prompt injection and jailbreak detector. Filters malicious prompts and third-party data for LLM apps. 99%+…

akillness/ai-research-skills

1

13

pytorch-fsdp2 Adds PyTorch FSDP2 (fully_shard) to training scripts with correct init, sharding, mixed precision/offload config, and…

akillness/ai-research-skills

1

14

pytorch-lightning High-level PyTorch framework with Trainer class, automatic distributed training (DDP/FSDP/DeepSpeed), callbacks syste…

akillness/ai-research-skills

1

15

pyvene-interventions Provides guidance for performing causal interventions on PyTorch models using pyvene's declarative intervention frame…

akillness/ai-research-skills

1

16

qdrant-vector-search High-performance vector similarity search engine for RAG and semantic search. Use when building production RAG system…

akillness/ai-research-skills

1

17

quantizing-models-bitsandbytes Quantizes LLMs to 8-bit or 4-bit for 50-75% memory reduction with minimal accuracy loss. Use when GPU memory is limit…

akillness/ai-research-skills

1

18

ray-data Scalable data processing for ML workloads. Streaming execution across CPU/GPU, supports Parquet/CSV/JSON/images. Inte…

akillness/ai-research-skills

1

19

ray-train Distributed training orchestration across clusters. Scales PyTorch/TensorFlow/HuggingFace from laptop to 1000s of nod…

akillness/ai-research-skills

1

20

rwkv-architecture RNN+Transformer hybrid with O(n) inference. Linear time, infinite context, no KV cache. Train like GPT (parallel), in…

akillness/ai-research-skills

1

21

segment-anything-model Foundation model for image segmentation with zero-shot transfer. Use when you need to segment any object in images us…

akillness/ai-research-skills

1

22

sentence-transformers Framework for state-of-the-art sentence, text, and image embeddings. Provides 5000+ pre-trained models for semantic s…

akillness/ai-research-skills

1

23

sentencepiece Language-independent tokenizer treating text as raw Unicode. Supports BPE and Unigram algorithms. Fast (50k sentences…

akillness/ai-research-skills

1

24

serving-llms-vllm Serves LLMs with high throughput using vLLM's PagedAttention and continuous batching. Use when deploying production L…

akillness/ai-research-skills

1

25

sglang Fast structured generation and serving for LLMs with RadixAttention prefix caching. Use for JSON/regex outputs, const…

akillness/ai-research-skills

1

26

simpo-training Simple Preference Optimization for LLM alignment. Reference-free alternative to DPO with better performance (+6.4 poi…

akillness/ai-research-skills

1

27

skypilot-multi-cloud-orchestration Multi-cloud orchestration for ML workloads with automatic cost optimization. Use when you need to run training or bat…

akillness/ai-research-skills

1

28

slime-rl-training Provides guidance for LLM post-training with RL using slime, a Megatron+SGLang framework. Use when training GLM model…

akillness/ai-research-skills

1

29

sparse-autoencoder-training Provides guidance for training and analyzing Sparse Autoencoders (SAEs) using SAELens to decompose neural network act…

akillness/ai-research-skills

1

30

speculative-decoding Accelerate LLM inference using speculative decoding, Medusa multiple heads, and lookahead decoding techniques. Use wh…

akillness/ai-research-skills

1

31

stable-diffusion-image-generation State-of-the-art text-to-image generation with Stable Diffusion models via HuggingFace Diffusers. Use when generating…

akillness/ai-research-skills

1

32

tensorboard Visualize training metrics, debug models with histograms, compare experiments, visualize model graphs, and profile pe…

akillness/ai-research-skills

1

33

tensorrt-llm Optimizes LLM inference with NVIDIA TensorRT for maximum throughput and lowest latency. Use for production deployment…

akillness/ai-research-skills

1

34

torchforge-rl-training Provides guidance for PyTorch-native agentic RL using torchforge, Meta's library separating infra from algorithms. Us…

akillness/ai-research-skills

1

35

training-llms-megatron Trains large language models (2B-462B parameters) using NVIDIA Megatron-Core with advanced parallelism strategies. Us…

akillness/ai-research-skills

1

36

transformer-lens-interpretability Provides guidance for mechanistic interpretability research using TransformerLens to inspect and manipulate transform…

akillness/ai-research-skills

1

37

unsloth Expert guidance for fast fine-tuning with Unsloth - 2-5x faster training, 50-80% less memory, LoRA/QLoRA optimization

akillness/ai-research-skills

1

38

verl-rl-training Provides guidance for training LLMs with reinforcement learning using verl (Volcano Engine RL). Use when implementing…

akillness/ai-research-skills

1

39

weights-and-biases Track ML experiments with automatic logging, visualize training in real-time, optimize hyperparameters with sweeps, a…

akillness/ai-research-skills

1

40

whisper OpenAI's general-purpose speech recognition model. Supports 99 languages, transcription, translation to English, and …

akillness/ai-research-skills

1